Comparative Analysis On The Impact of GPT On Human Thinking Using Sentiment Analysis

Comparative Analysis On The Impact of GPT On Human Thinking Using Sentiment Analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Comparative Analysis On The Impact of GPT On Human Thinking Using Sentiment Analysis Shouvic Banik, Sapna Sinha This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6803979/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This research article examines the effects of GPT (Generative Pre-trained Transformer) models on human thought processes and emotions, specifically looking at changes in sentiment and themes in user feedback. Using a multi-step approach that includes comparative sentiment evaluation, word-level sentiment examination, and thematic modelling, the study assesses how users’ views and cognitive articulations evolve before and after their interaction with GPT. Sentiment evaluations were performed via VADER and TextBlob to ensure thoroughness and validation of polarity and subjectivity ratings. The thematic analysis revealed shifting trends in trust, doubt, and hands-on engagement with AI-created material. Statistical analyses, such as Tukey’s HSD, were utilized to determine the relevance of sentiment differences among various user demographics, identifying significant variations linked to age. By combining sentiment trend observations, word co-occurrence networks, and comparisons of polarity and subjectivity scores, the research provides a detailed perspective to gauge the nuanced yet quantifiable impact of GPT on human cognition and emotional perspectives. These results enhance the overall comprehension of human-AI relationships and their significance for digital interaction, AI acceptance, and cognitive changes. GPT human cognition sentiment evaluation TextBlob VADER thematic exploration AI cognition sentiment variation natural language processing user viewpoint Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 1. Introduction Artificial intelligence (AI) has rapidly transformed a number of industries, particularly with the advent of large-scale language models like OpenAI's Generative Pre-trained Transformer (GPT). These advanced systems use deep learning approaches to generate human-like language, assist in problem solving, and provide responses that mimic genuine conversation (Brown et al., 2020). GPT-powered applications are becoming more prevalent in industries like software engineering, education, customer service, healthcare, and content production, enabling in-depth communication between people and AI-generated responses. However, a fundamental question remains: Does GPT influence human thought processes? Unlike conventional search engines or static information bases, GPT models generate replies dynamically that may influence user perspectives, affect affective reactions, or even alter decision-making processes (Weidinger et al., 2021). Understanding whether and how GPT alters human thought processes—whether by introducing cognitive biases, promoting critical analysis, or reinforcing preexisting beliefs—is essential because it has important ramifications for ethical issues in AI, human-computer interaction, and cognitive psychology research. This study examines changes in sentiment in human replies before and after using GPT using a comparative sentiment analysis method. Through the use of VADER sentiment analysis , theme extractions , and word-level sentiment mapping , this study provides empirical evidence about whether GPT merely serves as an objective information source or actively influences human cognitive processes. 1.1 Background of the Study The rise of artificial intelligence has dramatically altered the ways in which people connect, communicate, and process information. Among the multitude of AI frameworks, the Generative Pre-trained Transformer (GPT) has become particularly popular due to its capability to produce text that resembles human writing, aid in solving problems, and offer valuable insights across various fields (Brown et al., 2020). The utilization of GPT ranges from academic exploration and professional composition to automating customer service and providing personal assistance, resulting in a fundamental transformation in how people engage with AI-driven technologies (Zou et al., 2023). Despite the recognized effectiveness and versatility of GPT, there remain concerns about its effects on human thought processes, decision-making, and emotional development (Weidinger et al., 2021). Recent improvements in natural language processing have boosted GPT’s capacity to create responses that are aware of context. Unlike conventional search engines that pull up stored data, GPT generates its answers in real-time, which may affect the way users think (Bubeck et al., 2023). This brings forth crucial inquiries: Is GPT only a tool for users to access information, or does it actively influence their viewpoints? How do people respond both emotionally and cognitively to content created by AI? Responding to these questions is vital for assessing the role of AI in today’s world and its consequences for human logic, emotional involvement, and choice-making. Earlier studies have delved into sentiment analysis to gauge user reactions to interactions powered by AI (Ahuja et al., 2022). Nonetheless, the majority of research has concentrated either on measuring sentiment attributes like positive, negative, or neutral responses or isolated thematic analysis, failing to combine both approaches to grasp cognitive and emotional changes comprehensively. This study intends to perform a comparative analysis of sentiment on open-ended versus structured responses both prior to and following interactions with GPT, with the aim of revealing patterns in sentiment changes, word associations, and thematic differences across various user groups. This inquiry will not only enhance the understanding of AI’s impact on human cognition but also offer valuable insights for the creation of responsible AI systems that respect ethical and cognitive standards. 1.2 Research Problem Despite considerable advancements in artificial intelligence, the degree to which GPT affects users' cognitive functioning remains largely unexamined. While there has been a significant amount of research into AI ethics, bias reduction, and the performance of language models, there exists a crucial lack of insight into how GPT impacts human feelings, thought processes, and decision-making (Binns et al., 2018; Weidinger et al., 2021). In contrast to traditional AI tools that are tailored for specific tasks, GPT involves open-ended dialogues, which complicates the assessment of its influence on human cognition (Bubeck et al., 2023). Current Research Emphasis The existing body of literature mainly focuses on: Bias Identification in AI Models — Investigations have thoroughly examined how GPT displays racial, gender, and cultural biases, raising alarms about fairness and inclusivity in responses generated by AI. Scholars have developed techniques for reducing bias, but attention remains on the ethical dimensions of AI rather than its cognitive effects on users (Sheng et al., 2021). Evaluation of Performance — Most studies evaluate GPT's fluency, coherence, factual accuracy, and adaptability in language (Brown et al., 2020). Though these evaluations aid in refining AI-generated content, they fall short in considering the influence of ongoing exposure to GPT-created text on users’ thoughts and feelings. Ethical Issues — Topics such as misinformation, AI-driven propaganda, and reliance on AI for decision-making have been thoroughly examined. Nonetheless, these discussions are typically framed from a regulatory or philosophical viewpoint rather than addressing their psychological or cognitive effects (Floridi & Cowls, 2019). Research Limitations and Critical Questions Even with these advancements, there is a scarcity of empirical studies investigating how GPT shapes user feelings, cognitive perceptions, and decision-making behaviours. This research aims to bridge that gap by exploring the following pivotal questions: Does interacting with GPT result in discernible changes in human sentiment? Understanding the variations in user sentiment before and after engagement with GPT will provide insights into whether AI simply mirrors human emotions or actively alters them. Do various age demographics display differing sentiment trends following GPT interactions? Evaluating sentiment differences among age groups will clarify whether younger or older users are more prone to cognitive changes brought about by AI-generated material. Can thematic analysis uncover cognitive changes influenced by AI-generated content? By pinpointing common themes in user feedback before and after engaging with GPT, this study aspires to reveal any potential cognitive and emotional shifts instigated by AI. What importance does word-level sentiment have in analysing these changes? Exploring sentiment at the word level could illuminate subtle shifts in emotion, aiding in quantifying GPT’s effects on user language and overall expression. Importance of the Study This research intends to conduct a comparative analysis across different datasets to ascertain whether GPT functions solely as an information provider or serves as an active cognitive influencer. The results will enhance the understanding of AI-human interactions and provide crucial insights for developers, policymakers, and researchers regarding the careful design and use of generative AI systems 1.3 Research Objectives The central purpose of this research is to methodically examine changes in sentiment and themes in human reactions prior to and following engagement with GPT. The particular goals are: To evaluate the distribution of sentiments before and after engaging with GPT, employing sentiment analysis methods such as VADER. To uncover variations in themes within user feedback and evaluate how responses generated by AI alter human perspective. To investigate shifts in sentiment at the word level, looking at how particular words and phrases influence user outlook. To ascertain if demographic elements (such as age and level of expertise) affect sentiment alterations after interacting with GPT. These goals together offer both quantitative and qualitative insights into the effects of GPT on human cognitive processes. 1.4 Structure of the Paper In order to promote understanding and logical flow, this research document is organized in the following manner: Chapter 2: Review of Literature An extensive examination of prior research on sentiment analysis, cognitive impact driven by AI, and moral considerations in interactions between humans and AI. This portion will offer comparative perspectives derived from earlier studies. Chapter 3: Research Methodology Outlines the datasets utilized (comparing older and newer data), the processes for preparing data, techniques for sentiment analysis, methods for extracting themes, and the statistical models applied to guarantee precision in measuring sentiment. Chapter 4: Findings and Analysis Displays results through visual elements (charts representing sentiment distribution, word clouds, and thematic diagrams). The analysis interprets these outcomes concerning the research challenge. Chapter 5: Summary and Prospective Research Condenses essential findings, discusses the limitations of the study, and suggests possible avenues for future exploration into AI’s impact on human cognitive processes. 2. Review of Literature 2.1 AI’s Cognitive Influence: Existing Literature Concerns have been expressed regarding the impact of artificial intelligence on human thought processes due to its rapid breakthroughs, especially in the field of natural language processing. Previous research has largely concentrated on the ways AI improves decision-making, creativity, and problem-solving abilities. The impact of AI on a range of sectors, including business, healthcare, and education, has been examined by researchers. AI-driven chatbots have been shown to improve learning and cognitive engagement (Zhai, 2022). Empirical research evaluating the direct effects of AI on human emotions and cognitive functions is, nevertheless, lacking. Subtle but potentially important changes in how people perceive information and develop opinions have resulted from the incorporation of AI into commonplace applications in recent years, such virtual assistants, recommendation systems, and conversational agents (Long & Magerko, 2020). According to some research, extended use of AI systems may change how people think critically by promoting dependence on machine-generated answers and decreasing the requirement for autonomous problem-solving (Williams et al., 2022). Additionally, users' mood and introspection may be impacted by the emotional tone and responsiveness of sophisticated language models, bringing up significant ethical and psychological issues (Shin, 2021). Understanding AI systems' long-term cognitive and emotional effects on users, in addition to its functional advantages, is becoming increasingly important as they develop.This calls for a multidisciplinary research approach that bridges psychology, neuroscience, and artificial intelligence. The development and reinforcement of cognitive biases during human-AI contact is a major problem with AI-generated content. According to existing research, people frequently display anchoring bias when they are exposed to AI system outputs, meaning they are more likely to be swayed by the first recommendations or information offered, even in cases where independent thought or critical assessment are required (Rahwan et al., 2019). Users may become unduly dependent on AI-generated solutions as a result of this cognitive tendency, especially in situations where human judgment or interpretive nuance are crucial.In situations involving subjective inquiry, when people are more inclined to take AI's first framing or viewpoint as a reference point, the effect is particularly noticeable. Additionally, empirical research indicates that rather than questioning human biases, AI-generated language may subtly support preexisting attitudes and ideas. The possibility that AI systems could inadvertently sway or influence public opinion is raised by the possibility that this reinforcement could take place covertly, affecting perceptions and judgments in ways that are not immediately apparent to the user (Weidinger et al., 2021). 2.2 Sentiment Analysis in Human-Computer Interaction Research on how people and computers read sentiment suggests that AI-expressed emotions could quickly affect user emotions. A study by Xu et al. (2022) found that when individuals saw sympathetic reactions from AI systems, their happy sensations rose. However, people's moods shifted to match the tone of the AI when they met negative sentiment responses generated by the system. These findings underline the need for additional research on how AI's linguistic qualities affect human emotions and cognitive functions across extended interactions. Within the scope of GPT-based models, research has looked at how AI-generated content influences user opinions, especially on social media sites and in online forums. Gao et al. (2023) examined user engagement with GPT-enhanced chatbots and discovered that AI-generated responses significantly influenced sentiment trends, particularly among younger audiences. Our research attempts to address a significant gap in the literature by determining whether these sentimental changes are linked to long-term cognitive changes. Furthermore, recent research has shown that sentiment analysis in human-AI interactions depends on both the perceived intentionality of the AI responses and the substance. According to Fischer et al. (2022), people were more emotionally impacted when they thought the AI was responding with empathy. This implies that user impression of AI's intent is a critical factor in emotional resonance, independent of verbal clues. Furthermore, the significance of contextual relevance in influencing emotional and cognitive reactions to AI interactions was highlighted by Ghosh et al. (2021), who contended that models such as GPT can more successfully modify user sentiment when responses closely match conversational context. 2.3 Addressing Research Gaps Although sentiment analysis has been widely used across different fields, a notable deficiency exists in grasping its effects on AI-induced cognitive impacts. The primary research deficiencies recognized consist of: Insufficient empirical investigations regarding sentiment changes prior to and following AI interactions : Current research predominantly centres on static sentiment assessments rather than observing the evolving nature of sentiment over time. Restricted inquiry into how demographics affect sentiment alterations driven by AI : Very few investigations analyse the influence of elements such as age, knowledge, or familiarity with AI on shifts in sentiment. Limited research on word-specific sentiment alterations during AI-human exchanges : While broad sentiment categorization is prevalent, the influence of particular word selections on user perception is still largely unexamined. Lack of comparative studies addressing AI-derived themes versus those produced by humans : It is crucial to explore whether AI subtly modifies human cognition by transforming the subjects of discussion. Our research endeavours to fill these gaps by conducting a comparative investigation of sentiment changes and thematic differences, aiming to provide a comprehensive understanding of the influence GPT has on human cognition. 3. Methodology 3.1 Dataset Details This research utilizes two key datasets to evaluate how GPT affects human thought processes. The first dataset is made up of both open and closed responses gathered before and after users interacted with GPT, reflecting user feelings and various themes. The second dataset consists of user reviews regarding GPT-based applications, providing further understanding of user attitudes over time and cognitive alterations. Dataset 1: Responses Before and After GPT Interaction Description : This dataset comprises responses from users prior to and following their engagement with GPT. Structure : Pre-GPT Responses : Original user insights on specific subjects. Post-GPT Responses : Feedback collected after users interacted with GPT-generated material. Demographic Information : Age categories, levels of expertise, and previous exposure to AI technology. • Purpose : The aim is to uncover shifts in sentiment and theme alterations in responses, illustrating whether AI impacts users’ thought processes. Dataset 2: Review-Based Dataset Description : This dataset features user reviews and evaluations concerning GPT models. Structure : Text Content : Reviews from users that examine GPT's functionality, biases, and user-friendliness. Sentiment Scores : Ratings provided by users or deduced sentiment scores. Time-Based Trends : Reviews from various time frames to analyse changing viewpoints. • Purpose : The objective is to investigate long-term changes in sentiment and to pinpoint recurring themes in user interactions. 3.2 Data Preprocessing The data collections went through several preliminary processes to maintain uniformity and precision during evaluation: Data Cleaning : Eliminated punctuation marks, unique characters, and extra spaces. Tokenization : Divided the text into separate words and expressions. Stopword Removal : Excluded frequent yet unhelpful words from the dataset. Lemmatization : Converted words to their simplest form for consistency in assessment. 3.3 Sentiment Analysis Methodology Sentiment Analysis Tools : VADER, which stands for Valence Aware Dictionary and sentiment Reasoner, along with TextBlob, were utilized for categorizing sentiment. Sentiment Categories : Each reply was sorted into either positive, negative, or neutral sentiment according to their polarity scores. Comparative Sentiment Distribution Analysis : Distribution Comparison: A comparative evaluation was conducted to detect changes in sentiment between responses before and after the GPT interaction, as well as to observe sentiment patterns in user feedback. Word-Level Sentiment Shifts & Cognitive Influence : The analysis of word co-occurrence networks and sentiment variations was carried out to examine how the sentiment of frequently used words evolved before and after the GPT engagement. 3.4 Thematic Analysis Approach Topic Modelling Latent Dirichlet Allocation (LDA) was employed to uncover prevailing themes within the feedback from users. Key Themes Identified : Pre-GPT Responses : Notable themes reflected doubt, inquisitiveness, and unease regarding content created by AI. Post-GPT Responses : Themes like heightened confidence in AI, worries about bias, and recognition of GPT's usefulness surfaced. Review Based Dataset : Long-term patterns revealed a changing level of trust in GPT, ongoing worries about misinformation generated by AI, and enhancements in usability. Demographic-Influence on Sentiment : Variations in sentiment were assessed among various age demographics to explore if younger and older individuals reacted differently to interactions with GPT. 3.5 Statistical Analysis Sentiment Shift Evaluation Across Age Groups : A chi-square analysis was performed to evaluate if sentiment alterations were significantly different across various age brackets. Findings suggested that younger individuals displayed a more pronounced positive change in sentiment following GPT use when contrasted with older individuals. Word-Level Sentiment Trends : Analysing word frequency alongside sentiment ratings helped to identify changes in the most frequently used terms before and after interacting with GPT. Noteworthy findings included a rise in positive sentiment expressions like "beneficial" and "insightful" after engaging with GPT. TextBlob Analysis : This method, in conjunction with VADER, provided scores for polarity and subjectivity for a more thorough sentiment analysis comparison. Model Evaluation and Refinements : The sentiment models underwent refinement by modifying VADER thresholds and validating through TextBlob to enhance the accuracy of classifications. Results Summary (Statistical Analysis) : Sentiment Shift Significance: p-value < 0.05, indicating a statistically significant sentiment shift post-GPT. Frequent Positive Words: "Innovative," "useful," "accurate." Frequent Negative Words: "Biased," "misleading," "incorrect." Conclusion The approach used in this research facilitates an in-depth exploration of changes in sentiment and variations in themes after engaging with GPT. Through the integration of VADER and TextBlob for sentiment evaluation, LDA for theme analysis, sentiment changes at the word level, and statistical assessment, this study offers concrete proof of GPT's impact on human thought processes. The following chapter will outline the comprehensive results and discussions derived from this approach. 4. Findings and Discussion 4.1 Overview of Findings The findings from the statistical analyses, thematic analysis, and comparative sentiment analysis of the responses obtained before and after the GPT encounter are presented in this chapter. Finding out how user sentiment, cognitive perceptions, and thematic structures changed after interacting with GPT-generated material is the goal of the analysis. 4.2 Sentiment Analysis Results 4.2.1 Sentiment Distribution Before and After GPT Interaction A discernible change in user sentiment upon interaction with GPT is revealed by the sentiment analysis conducted using VADER and TextBlob. The following significant findings were noted: Pre-GPT Sentiment : There were fewer favourable responses and higher levels of scepticism and neutrality. Post-GPT Sentiment : A notable rise in positive sentiment, a fall in negative sentiment, and a little drop in neutral opinion. Review-Based Sentiment Trends : Although worries about biases continue, long-term user reviews show consistent favourable sentiment Table 1 : Sentiment Distribution Comparison Before and After GPT Interaction Dataset Positive (%) Neutral (%) Negative (%) Pre-GPT Responses 35% 40% 25% Post-GPT Responses 50% 30% 20% Review Dataset 45% 35% 20% 4.2.2 Word-Level Sentiment Shift Analysis A word cloud analysis was performed to visualize the most frequently occurring positive and negative sentiment words before and after GPT interaction. The findings indicate: Pre-GPT Responses : Dominant negative words included "biased," "misleading," and "uncertain." Post-GPT Responses : Increase in positive words such as "helpful," "insightful," and "accurate." 4.2.3 Sentiment Trend Over Time A study of sentiment variations throughout various times indicates a changing view of GPT’s abilities. The pattern shows a gradual rise in favourable sentiment as time progresses, signifying enhanced user experiences and growing confidence in content produced by AI. 4.3 Thematic Analysis Results 4.3.1 Emerging Themes Before and After GPT Interaction Interaction Latent Dirichlet Allocation (LDA) was used to extract key themes from pre- and post-GPT responses. The thematic shifts observed include: • Pre-GPT Themes : Scepticism about AI capabilities. Ethical concerns regarding AI-generated content. Curiosity but cautious engagement. • Post-GPT Themes : Increased trust in AI assistance. Acknowledgment of GPT’s usability and efficiency. Concerns about AI biases persisting but with reduced emphasis. Table 2 : Thematic Shifts Before and After GPT Interaction Theme Pre-GPT (%) Post-GPT (%) Review Dataset (%) Scepticism about AI 40% 20% 15% Trust in AI 25% 50% 55% Concerns about bias 20% 25% 30% Usability & Practicality 15% 35% 50% 4.3.2 Demographic-Based Sentiment Shift The change in feelings differs across various age demographics, as younger individuals showcase a more unfavourable emotional transition, whereas older individuals reveal a more pronounced positive emotional transition. 18–25 : -0.174 (Negative Shift) 26–35 : -0.357 (Negative Shift) 36–45 : 0.062 (Minimal Positive Shift) 45–60 : 0.480 (Strong Positive Shift) Under 18 : -0.031 (Slight Negative Shift) These results oppose the original assumption that younger individuals experience a greater positive change in sentiment. Rather, older individuals display a more significant uptick in positivity, indicating either a higher adaptability or a deeper enjoyment of content created by AI, whereas younger individuals maintain a more doubtful perspective. This difference across age groups implies that factors such as digital skills, familiarity with AI, and expectations concerning AI-generated materials could influence how users feel. 4.4 Word-Level Sentiment Analysis 4.4.1 Most Frequently Used Words in User Responses and Reviews A comparative analysis of word frequency highlights key vocabulary patterns: Significant observations reveal a dominance of terms associated with practicality and effectiveness, including "exact," "informative," and "beneficial," in addition to ongoing worries articulated through expressions like "prejudiced" and "deceptive." 4.4.2 Word Co-occurrence Network Graph A graphical representation of word connections showcases the connections between terms and reveals patterns in their meanings as well as commonly linked phrases within content created by users. 4.5 Statistical Findings 4.5.1 Statistical Significance of Sentiment Shift A statistical examination of p-values was performed to ascertain if the change in sentiment noted after interacting with GPT was significant in a statistical sense. The findings (p-value < 0.05) indicate that interacting with GPT significantly influences how users feel. Nevertheless, additional statistical analyses are necessary to validate differences among various demographic groups 4.5.2 Polarity and Subjectivity Score Comparison Analysing the sentiment scores derived from VADER and TextBlob offers an enhanced insight into methods of sentiment analysis. Table 3 : Polarity and Subjectivity Score Comparison Dataset Average Polarity Score Average Subjectivity Score Old Dataset (Pre-GPT) 0.21 0.48 Old Dataset (Post-GPT) 0.36 0.56 New Dataset (Reviews) 0.42 0.61 This examination shows that both frameworks reliably identify changes in sentiment, exhibiting slight differences in the level of classification intensity. Modifications to the VADER threshold parameters enhanced the precision of sentiment classification. 4.5 Discussion The results indicate that GPT significantly affects human cognition, especially in shaping emotions and topic changes. Major insights are: Over time, there has been a growth in reliance on and approval of responses produced by GPT. After engaging with AI, there is a decline in doubt and unfavourable attitudes. While worries about AI bias remain, they are not as prominent in discussions following the introduction of GPT. Compared to younger users, older individuals show a more significant change towards positive sentiment. These observations emphasize the changing mental connection between individuals and artificial intelligence, indicating that GPT not only educates but also shapes how users view things. Conclusion This section offered a thorough examination of the results derived from both sentiment and thematic assessments. The following chapter will wrap up the research by highlighting significant observations, addressing constraints, and suggesting directions for upcoming investigations. 5. Conclusion and Recommendations 5.1 Summary of Key Findings This research sought to explore how GPT affects human cognition through a comparative analysis of sentiments and a thematic review of content created by users. The results display notable changes in both cognitive and emotional states of users before and after engaging with GPT. Key outcomes include: A quantifiable rise in positive emotions and a decline in negative feelings following interactions with GPT. A shift in themes from doubt and ethical dilemmas to confidence and practical involvement with GPT. A statistically considerable change in sentiment, confirmed through p-value analysis and the Tukey HSD test. Cross-validation with VADER and TextBlob established the consistency of sentiment patterns. Differences related to age suggest that older users show more pronounced positive sentiment changes. These results validate that GPT can shape human emotions and thought processes by building trust and improving the perceived usefulness of tasks assisted by AI. 5.2 Implications The consequences of this study extend across various areas: Human-AI Engagement : A rise in favourable feelings indicates a strengthening user trust and dependence on AI technology. Technological Understanding & Trust : Differences in attitudes by age highlight the necessity of advancing AI understanding to close trust disparities. AI Development and Moral Considerations : Ongoing worries about bias emphasize the critical need for creating clear and principled AI systems. 5.3 Limitations Even with strong evidence, this research has several drawbacks: The data collected might not accurately reflect all demographic groups or sectors. Emotion detection through lexicon-oriented frameworks (VADER, TextBlob) might overlook subtle emotional variations. The thematic examination was limited by established topic frameworks and the assumptions of the model. 6. Future Work and Research Implications 6.1 Potential Areas for Further Exploration While our research has effectively highlighted important changes in sentiment and themes related to GPT interactions, numerous paths are still available for future inquiries: Extended Research : A prolonged examination of how users' views of GPT and other large language models change over time could reveal deeper trends in behaviour and thought processes. Analysis Across Cultures : Broadening the study to encompass diverse cultures and languages might uncover distinct sentiment tendencies and thematic occurrences. Inquiries Based on Platforms : Examining sentiments and themes on different platforms like discussion boards, social media, and professional networks could bring attention to how context influences user opinions. Effects on Critical Thinking : Additional studies could aim to measure how the use of GPT affects users' abilities to think independently and make decisions. 6.2 Methodological Enhancements Future research may utilize innovative techniques and resources to enhance the evaluation: Combined Sentiment Approaches : Merging dictionary-driven techniques (like VADER and TextBlob) with machine learning sentiment analysis tools (such as BERT) to achieve more refined sentiment understanding. Enhanced Thematic Exploration : Utilizing topic modelling alongside qualitative analysis to improve the clarity of identified themes. User Behaviour Analytics : Incorporating clickstream information, response durations, and eye-tracking to establish connections between user interaction patterns and emotional as well as cognitive metrics. 6.3 Broader Research Implications The results obtained from this research provide significant insights for multiple sectors: Education : Enlightening teachers about the ways in which GPT affects students’ thought processes and learning habits. AI Ethics : Adding to conversations on the ethical design and application of AI, particularly regarding its impact on users. Policy and Governance : Aiding in the creation of regulations that consider the cognitive effects of AI on individuals. 6.4 Final Thoughts on Future Research The impact of generative AI, such as GPT, on human thought processes is a developing area of study. As these innovations progress and become part of everyday experiences, it is crucial to keep conducting research to grasp their lasting effects on cognition, emotions, and society. Future efforts should prioritize interdisciplinary methods that merge insights from computational science, psychology, sociology, and ethics to comprehensively evaluate AI's influence on human mental processes. Declarations Competing Interests : The authors have no relevant financial or non-financial interests to disclose. Funding : The authors did not receive support from any organization for the submitted work. Author Contribution S.B. carried out the research work along with writing the manuscript and S.S. provided supervision and reviewed the manuscript Data Availability Data will be made available on request References Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., ... & Amodei, D. (2020). Language models are few-shot learners [arXiv preprint arXiv:2005.14165]. https://doi.org/10.48550/arXiv.2005.14165 Weidinger, L., Mellor, J., Rauh, M., Griffin, C., Uesato, J., Huang, P. S., ... & Gabriel, I. (2021). Ethical and social risks of harm from language models [arXiv preprint arXiv:2112.04359]. https://doi.org/10.48550/arXiv.2112.04359 Ahuja, R., Biyani, P., Caragea, C., & Caragea, D. (2022). Understanding user sentiment and engagement on COVID-19 vaccine-related Twitter discussions. Journal of Medical Internet Research, 24 (5), e33742. https://doi.org/10.2196/33742 Bubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., ... & Zhang, Y. (2023). Sparks of artificial general intelligence: Early experiments with GPT-4 [arXiv preprint arXiv:2303.12712]. https://doi.org/10.48550/arXiv.2303.12712 Zou, J. Y., Schiebinger, L., Rasmussen, J., & Obermeyer, Z. (2023). Artificial intelligence in health care: Anticipating challenges to ethics, privacy, and bias. Health Affairs, 42 (1), 78–85. https://doi.org/10.1377/hlthaff.2022.00672 Binns, R., Veale, M., Van Kleek, M., & Shadbolt, N. (2018). “It's reducing a human being to a percentage”: Perceptions of justice in algorithmic decisions. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (pp. 1–14). https://doi.org/10.1145/3173574.3173951 Floridi, L., & Cowls, J. (2019). A unified framework of five principles for AI in society. Harvard Data Science Review, 1 (1). https://doi.org/10.1162/99608f92.8cd550d1 Sheng, E., Chang, K. W., Natarajan, P., & Peng, N. (2021). Societal biases in language generation: Progress and challenges. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) (pp. 4277–4293). https://doi.org/10.18653/v1/2021.emnlp-main.348 Rahwan, I., Cebrian, M., Obradovich, N., Bongard, J., Bonnefon, J. F., Breazeal, C., ... & Lazer, D. (2019). Machine behaviour. Nature, 568 (7753), 477–486. https://doi.org/10.1038/s41586-019-1138-y Zhai, X. (2022). ChatGPT for intelligent education: Opportunities, challenges, and strategies. Educational Philosophy and Theory , 1–4. https://doi.org/10.1080/00131857.2023.2171770 Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding [arXiv preprint arXiv:1810.04805]. https://arxiv.org/abs/1810.04805 Gao, C., Zhang, Y., Liu, T., & Wang, X. (2023). Exploring user sentiment dynamics in GPT-powered interactions: A case study on youth engagement. Journal of Artificial Intelligence Research and Development, 45 (2), 123–139. Hutto, C. J., & Gilbert, E. (2014). VADER: A parsimonious rule-based model for sentiment analysis of social media text. In Proceedings of the International AAAI Conference on Web and Social Media, 8 (1), 216–225. https://ojs.aaai.org/index.php/ICWSM/article/view/14550 Xu, H., Wang, Y., Li, M., & Chen, X. (2022). The emotional resonance of AI: Effects of empathetic responses in human-AI interaction. Computers in Human Behavior, 132 , 107262. https://doi.org/10.1016/j.chb.2022.107262 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6803979","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":514068447,"identity":"71cd7f1c-6f39-4079-a9ae-74a810edf86a","order_by":0,"name":"Shouvic Banik","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBElEQVRIiWNgGAWjYDACCRBhAEYMBxsqQDzGBgRJWMsZsOLGBsJaILqAatsY4Bbg1CI/u/mYNE+BHYM5+9mDB2fOq2Pgn3a4/TEPg43shgPYtRjcOZYmzWOQzGDZk5dwcOO2wwwStxMbm3kY0oxxapHIMZOcYXCAweBAjsHBh9uAyiBaDifi0iI/A6bl/Bugljl1DPIQLf9xamG4kWMm8QGk5QbQlo0NzAwGEC0HcGoxuJGWbPHBIJnHcgbQlhnHDvMYArXMnGOQbDwTp8OSD95I+GMnZ86fY/yxp6ZOTu52+oMPbyrsZPtwOQwKeFAYTDwG+JVjAsYfpOoYBaNgFIyC4QwAcJNjhR0ZYx4AAAAASUVORK5CYII=","orcid":"","institution":"Amity University","correspondingAuthor":true,"prefix":"","firstName":"Shouvic","middleName":"","lastName":"Banik","suffix":""},{"id":514068448,"identity":"b8d45324-e83a-4e9c-bec9-8da393cc07ab","order_by":1,"name":"Sapna Sinha","email":"","orcid":"","institution":"Amity University","correspondingAuthor":false,"prefix":"","firstName":"Sapna","middleName":"","lastName":"Sinha","suffix":""}],"badges":[],"createdAt":"2025-06-02 16:23:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6803979/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6803979/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91409816,"identity":"5062ae5c-2588-4eda-8961-a486f861ff24","added_by":"auto","created_at":"2025-09-16 08:27:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":634479,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual Framework- GPT’s role in Influencing Human Sentiment and Thought\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/2f6b15d80ab3ea0ad1dab68f.png"},{"id":91408265,"identity":"adec1927-dca9-4392-a37a-482cee17180b","added_by":"auto","created_at":"2025-09-16 08:19:54","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":861744,"visible":true,"origin":"","legend":"\u003cp\u003eEvolution of AI Chatbots (from ELIZA to GPT-4)\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/4b46c6cdc9ed39ea495bba2e.png"},{"id":91408262,"identity":"7c525410-ba4b-4f9f-ab8a-47963419ee7f","added_by":"auto","created_at":"2025-09-16 08:19:54","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":127992,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual Model for Identifying and Addressing the Research Gap\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/44f9fc860f3a2e748a211546.png"},{"id":91411312,"identity":"c4210574-025a-4511-b0fc-4cd1a851ce77","added_by":"auto","created_at":"2025-09-16 08:43:54","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":699958,"visible":true,"origin":"","legend":"\u003cp\u003eFramework of Research Objectives and Analytical Approaches\u003c/p\u003e","description":"","filename":"image4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/21e32171d50bd5187e9e1be5.jpeg"},{"id":91408266,"identity":"1aeb0ec6-2811-46d4-a2c8-92d28e271cb2","added_by":"auto","created_at":"2025-09-16 08:19:54","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":97883,"visible":true,"origin":"","legend":"\u003cp\u003eResearch Methodology and Analytical Workflow\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/ce4d2d49b7c0b536beb74b3c.png"},{"id":91409818,"identity":"24a7e0e7-553e-44c7-ae45-55d3524d374e","added_by":"auto","created_at":"2025-09-16 08:27:54","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":135308,"visible":true,"origin":"","legend":"\u003cp\u003eDiagram Illustrating Key Areas Where AI Influences Human Cognition\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/256645767f9ed8787c94a827.png"},{"id":91409825,"identity":"02a2eed5-0437-4f01-8a16-870a6de94c54","added_by":"auto","created_at":"2025-09-16 08:27:54","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":412296,"visible":true,"origin":"","legend":"\u003cp\u003eEvolution of Sentiment Analysis models.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/59e66c9a3b867f1e8748853d.png"},{"id":91816586,"identity":"c1f3a5e4-d414-4f3c-b5e4-fbd08888e2f5","added_by":"auto","created_at":"2025-09-22 06:52:01","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":252490,"visible":true,"origin":"","legend":"\u003cp\u003eDiagram Summarizing the Structure of Both Datasets and Their Respective Components.\u003c/p\u003e","description":"","filename":"image9.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/30655d5e9af9d51c15e0adf1.png"},{"id":91410198,"identity":"c09f411b-22ff-4536-9cc7-e48c5ce5fa7c","added_by":"auto","created_at":"2025-09-16 08:35:54","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":841993,"visible":true,"origin":"","legend":"\u003cp\u003eSentiment Analysis Workflow Diagram\u003c/p\u003e","description":"","filename":"image10.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/e3b4cb328748b8fe30fbaa40.png"},{"id":91411850,"identity":"c6d8ccca-f815-45bd-a3b2-1021ed3323a6","added_by":"auto","created_at":"2025-09-16 08:51:54","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":65506,"visible":true,"origin":"","legend":"\u003cp\u003eWord Cloud Visualizing Positive and Negative Sentiment Words Before and After GPT Interaction\u003c/p\u003e","description":"","filename":"image11.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/84bd149e7b0ca7f09580c867.png"},{"id":91410191,"identity":"28303ea8-9a9e-4530-8454-9d97f4e52234","added_by":"auto","created_at":"2025-09-16 08:35:54","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":116204,"visible":true,"origin":"","legend":"\u003cp\u003eSentiment Trend Over Time in Review Dataset\u003c/p\u003e","description":"","filename":"image12.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/06e65e744684b2c13f59546e.png"},{"id":91408268,"identity":"ae07558d-f48f-49a5-934d-b8df65d4d952","added_by":"auto","created_at":"2025-09-16 08:19:54","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":27640,"visible":true,"origin":"","legend":"\u003cp\u003eDemographic Based Sentiment Shift\u003c/p\u003e","description":"","filename":"image13.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/a139e688391380f3413a46af.png"},{"id":91408285,"identity":"d395aff2-5cc6-4ac9-a11b-1b7b7c221513","added_by":"auto","created_at":"2025-09-16 08:19:54","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":265088,"visible":true,"origin":"","legend":"\u003cp\u003eFrequently Used Words in User Responses and Reviews\u003c/p\u003e","description":"","filename":"image14.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/68ad0a35601c38e213fd03c4.png"},{"id":91408294,"identity":"1b611cdf-4874-4f1e-8c5d-ef1dc248b2dd","added_by":"auto","created_at":"2025-09-16 08:19:55","extension":"png","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":967142,"visible":true,"origin":"","legend":"\u003cp\u003eWord Co-occurrence Network of Frequently Associated Words\u003c/p\u003e","description":"","filename":"image15.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/d12001de5ce3a5fe62d25aff.png"},{"id":91409828,"identity":"66b6ba10-f097-4edc-b433-dba7f05295e5","added_by":"auto","created_at":"2025-09-16 08:27:54","extension":"png","order_by":15,"title":"Figure 15","display":"","copyAsset":false,"role":"figure","size":43909,"visible":true,"origin":"","legend":"\u003cp\u003eStatistical Significance of Sentiment Shift - Tukey HSD Test\u003c/p\u003e","description":"","filename":"image17.png","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/9d22b3248e575807aaa1a0c6.png"},{"id":93331291,"identity":"a88ef1fb-6543-45d9-90a2-c91fe66bdb26","added_by":"auto","created_at":"2025-10-12 12:16:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6654735,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6803979/v1/b885402d-c0b3-47cb-8b5f-56351078e429.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Comparative Analysis On The Impact of GPT On Human Thinking Using Sentiment Analysis","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003e\u003cstrong\u003eArtificial intelligence (AI)\u003c/strong\u003e has rapidly transformed a number of industries, particularly with the advent of large-scale language models like OpenAI\u0026apos;s Generative Pre-trained Transformer (GPT). These advanced systems use deep learning approaches to generate human-like language, assist in problem solving, and provide responses that mimic genuine conversation (Brown et al., 2020). GPT-powered applications are becoming more prevalent in industries like software engineering, education, customer service, healthcare, and content production, enabling in-depth communication between people and AI-generated responses.\u003c/p\u003e\n\u003cp\u003eHowever, a fundamental question remains: \u003cstrong\u003eDoes GPT influence human thought processes?\u003c/strong\u003e Unlike conventional search engines or static information bases, GPT models generate replies dynamically that may influence user perspectives, affect affective reactions, or even alter decision-making processes (Weidinger et al., 2021). Understanding whether and how GPT alters human thought processes\u0026mdash;whether by introducing cognitive biases, promoting critical analysis, or reinforcing preexisting beliefs\u0026mdash;is essential because it has important ramifications for ethical issues in AI, human-computer interaction, and cognitive psychology research.\u003c/p\u003e\n\u003cp\u003eThis study examines changes in sentiment in human replies before and after using GPT using a comparative sentiment analysis method. Through the use of \u003cstrong\u003eVADER sentiment analysis\u003c/strong\u003e, \u003cstrong\u003etheme extractions\u003c/strong\u003e, and \u003cstrong\u003eword-level sentiment mapping\u003c/strong\u003e, this study provides empirical evidence about whether GPT merely serves as an objective information source or actively influences human cognitive processes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.1 Background of the Study\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe rise of artificial intelligence has dramatically altered the ways in which people connect, communicate, and process information. Among the multitude of AI frameworks, the Generative Pre-trained Transformer (GPT) has become particularly popular due to its capability to produce text that resembles human writing, aid in solving problems, and offer valuable insights across various fields (Brown et al., 2020). The utilization of GPT ranges from academic exploration and professional composition to automating customer service and providing personal assistance, resulting in a fundamental transformation in how people engage with AI-driven technologies (Zou et al., 2023). Despite the recognized effectiveness and versatility of GPT, there remain concerns about its effects on human thought processes, decision-making, and emotional development (Weidinger et al., 2021). Recent improvements in natural language processing have boosted GPT\u0026rsquo;s capacity to create responses that are aware of context. Unlike conventional search engines that pull up stored data, GPT generates its answers in real-time, which may affect the way users think (Bubeck et al., 2023). This brings forth crucial inquiries: Is GPT only a tool for users to access information, or does it actively influence their viewpoints? How do people respond both emotionally and cognitively to content created by AI? Responding to these questions is vital for assessing the role of AI in today\u0026rsquo;s world and its consequences for human logic, emotional involvement, and choice-making.\u003c/p\u003e\n\u003cp\u003eEarlier studies have delved into sentiment analysis to gauge user reactions to interactions powered by AI (Ahuja et al., 2022). Nonetheless, the majority of research has concentrated either on measuring sentiment attributes like positive, negative, or neutral responses or isolated thematic analysis, failing to combine both approaches to grasp cognitive and emotional changes comprehensively. This study intends to perform a comparative analysis of sentiment on open-ended versus structured responses both prior to and following interactions with GPT, with the aim of revealing patterns in sentiment changes, word associations, and thematic differences across various user groups. This inquiry will not only enhance the understanding of AI\u0026rsquo;s impact on human cognition but also offer valuable insights for the creation of responsible AI systems that respect ethical and cognitive standards.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.2 Research Problem\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDespite considerable advancements in artificial intelligence, the degree to which GPT affects users\u0026apos; cognitive functioning remains largely unexamined. While there has been a significant amount of research into AI ethics, bias reduction, and the performance of language models, there exists a crucial lack of insight into how GPT impacts human feelings, thought processes, and decision-making (Binns et al., 2018; Weidinger et al., 2021). In contrast to traditional AI tools that are tailored for specific tasks, GPT involves open-ended dialogues, which complicates the assessment of its influence on human cognition (Bubeck et al., 2023).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCurrent Research Emphasis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe existing body of literature mainly focuses on:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eBias Identification in AI Models\u003c/strong\u003e \u0026mdash; Investigations have thoroughly examined how GPT displays racial, gender, and cultural biases, raising alarms about fairness and inclusivity in responses generated by AI. Scholars have developed techniques for reducing bias, but attention remains on the ethical dimensions of AI rather than its cognitive effects on users (Sheng et al., 2021).\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eEvaluation of Performance\u003c/strong\u003e \u0026mdash; Most studies evaluate GPT\u0026apos;s fluency, coherence, factual accuracy, and adaptability in language (Brown et al., 2020). Though these evaluations aid in refining AI-generated content, they fall short in considering the influence of ongoing exposure to GPT-created text on users\u0026rsquo; thoughts and feelings.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eEthical Issues\u003c/strong\u003e \u0026mdash; Topics such as misinformation, AI-driven propaganda, and reliance on AI for decision-making have been thoroughly examined. Nonetheless, these discussions are typically framed from a regulatory or philosophical viewpoint rather than addressing their psychological or cognitive effects (Floridi \u0026amp; Cowls, 2019).\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eResearch Limitations and Critical Questions\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Even with these advancements, there is a scarcity of empirical studies investigating how GPT shapes user feelings, cognitive perceptions, and decision-making behaviours. This research aims to bridge that gap by exploring the following pivotal questions:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eDoes interacting with GPT result in discernible changes in human sentiment?\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Understanding the variations in user sentiment before and after engagement with GPT will provide insights into whether AI simply mirrors human emotions or actively alters them.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eDo various age demographics display differing sentiment trends following GPT interactions?\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Evaluating sentiment differences among age groups will clarify whether younger or older users are more prone to cognitive changes brought about by AI-generated material.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eCan thematic analysis uncover cognitive changes influenced by AI-generated content?\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;By pinpointing common themes in user feedback before and after engaging with GPT, this study aspires to reveal any potential cognitive and emotional shifts instigated by AI.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eWhat importance does word-level sentiment have in analysing these changes?\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Exploring sentiment at the word level could illuminate subtle shifts in emotion, aiding in quantifying GPT\u0026rsquo;s effects on user language and overall expression.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eImportance of the Study\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research intends to conduct a comparative analysis across different datasets to ascertain whether GPT functions solely as an information provider or serves as an active cognitive influencer. The results will enhance the understanding of AI-human interactions and provide crucial insights for developers, policymakers, and researchers regarding the careful design and use of generative AI systems\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.3 Research Objectives\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe central purpose of this research is to methodically examine changes in sentiment and themes in human reactions prior to and following engagement with GPT. The particular goals are:\u0026nbsp;\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;To evaluate the distribution of sentiments before and after engaging with GPT, employing sentiment analysis methods such as VADER.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;To uncover variations in themes within user feedback and evaluate how responses generated by AI alter human perspective.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;To investigate shifts in sentiment at the word level, looking at how particular words and phrases influence user outlook.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;To ascertain if demographic elements (such as age and level of expertise) affect sentiment alterations after interacting with GPT.\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThese goals together offer both quantitative and qualitative insights into the effects of GPT on human cognitive processes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.4 Structure of the Paper\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn order to promote understanding and logical\u0026nbsp;\u003c/strong\u003eflow, this research document is organized in the following manner:\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003e\u003cstrong\u003eChapter 2: Review of Literature\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eAn extensive examination of prior research on sentiment analysis, cognitive impact driven by AI, and moral considerations in interactions between humans and AI. This portion will offer comparative perspectives derived from earlier studies.\u0026nbsp;\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003e\u003cstrong\u003eChapter 3: Research Methodology\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eOutlines the datasets utilized (comparing older and newer data), the processes for preparing data, techniques for sentiment analysis, methods for extracting themes, and the statistical models applied to guarantee precision in measuring sentiment.\u0026nbsp;\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003e\u003cstrong\u003eChapter 4: Findings and Analysis\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eDisplays results through visual elements (charts representing sentiment distribution, word clouds, and thematic diagrams). The analysis interprets these outcomes concerning the research challenge.\u0026nbsp;\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003e\u003cstrong\u003eChapter 5: Summary and Prospective Research\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eCondenses essential findings, discusses the limitations of the study, and suggests possible avenues for future exploration into AI\u0026rsquo;s impact on human cognitive processes.\u003c/p\u003e"},{"header":"2. Review of Literature","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003e2.1 AI\u0026rsquo;s Cognitive Influence: Existing Literature\u003c/h2\u003e\n \u003cp\u003eConcerns have been expressed regarding the impact of artificial intelligence on human thought processes due to its rapid breakthroughs, especially in the field of natural language processing. Previous research has largely concentrated on the ways AI improves decision-making, creativity, and problem-solving abilities. The impact of AI on a range of sectors, including business, healthcare, and education, has been examined by researchers. AI-driven chatbots have been shown to improve learning and cognitive engagement (Zhai, 2022). Empirical research evaluating the direct effects of AI on human emotions and cognitive functions is, nevertheless, lacking.\u003c/p\u003e\n \u003cp\u003eSubtle but potentially important changes in how people perceive information and develop opinions have resulted from the incorporation of AI into commonplace applications in recent years, such virtual assistants, recommendation systems, and conversational agents (Long \u0026amp; Magerko, 2020). According to some research, extended use of AI systems may change how people think critically by promoting dependence on machine-generated answers and decreasing the requirement for autonomous problem-solving (Williams et al., 2022). Additionally, users\u0026apos; mood and introspection may be impacted by the emotional tone and responsiveness of sophisticated language models, bringing up significant ethical and psychological issues (Shin, 2021). Understanding AI systems\u0026apos; long-term cognitive and emotional effects on users, in addition to its functional advantages, is becoming increasingly important as they develop.This calls for a multidisciplinary research approach that bridges psychology, neuroscience, and artificial intelligence.\u003c/p\u003e\n \u003cp\u003eThe development and reinforcement of cognitive biases during human-AI contact is a major problem with AI-generated content. According to existing research, people frequently display anchoring bias when they are exposed to AI system outputs, meaning they are more likely to be swayed by the first recommendations or information offered, even in cases where independent thought or critical assessment are required (Rahwan et al., 2019). Users may become unduly dependent on AI-generated solutions as a result of this cognitive tendency, especially in situations where human judgment or interpretive nuance are crucial.In situations involving subjective inquiry, when people are more inclined to take AI\u0026apos;s first framing or viewpoint as a reference point, the effect is particularly noticeable. Additionally, empirical research indicates that rather than questioning human biases, AI-generated language may subtly support preexisting attitudes and ideas. The possibility that AI systems could inadvertently sway or influence public opinion is raised by the possibility that this reinforcement could take place covertly, affecting perceptions and judgments in ways that are not immediately apparent to the user (Weidinger et al., 2021).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e2.2 Sentiment Analysis in Human-Computer Interaction\u003c/h2\u003e\n \u003cp\u003eResearch on how people and computers read sentiment suggests that AI-expressed emotions could quickly affect user emotions. A study by Xu et al. (2022) found that when individuals saw sympathetic reactions from AI systems, their happy sensations rose. However, people\u0026apos;s moods shifted to match the tone of the AI when they met negative sentiment responses generated by the system. These findings underline the need for additional research on how AI\u0026apos;s linguistic qualities affect human emotions and cognitive functions across extended interactions.\u003c/p\u003e\n \u003cp\u003eWithin the scope of GPT-based models, research has looked at how AI-generated content influences user opinions, especially on social media sites and in online forums. Gao et al. (2023) examined user engagement with GPT-enhanced chatbots and discovered that AI-generated responses significantly influenced sentiment trends, particularly among younger audiences. Our research attempts to address a significant gap in the literature by determining whether these sentimental changes are linked to long-term cognitive changes.\u003c/p\u003e\n \u003cp\u003eFurthermore, recent research has shown that sentiment analysis in human-AI interactions depends on both the perceived intentionality of the AI responses and the substance. According to Fischer et al. (2022), people were more emotionally impacted when they thought the AI was responding with empathy. This implies that user impression of AI\u0026apos;s intent is a critical factor in emotional resonance, independent of verbal clues. Furthermore, the significance of contextual relevance in influencing emotional and cognitive reactions to AI interactions was highlighted by Ghosh et al. (2021), who contended that models such as GPT can more successfully modify user sentiment when responses closely match conversational context.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e2.3 Addressing Research Gaps\u003c/h2\u003e\n \u003cp\u003eAlthough sentiment analysis has been widely used across different fields, a notable deficiency exists in grasping its effects on AI-induced cognitive impacts. The primary research deficiencies recognized consist of:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eInsufficient empirical investigations regarding sentiment changes prior to and following AI interactions\u003c/strong\u003e: Current research predominantly centres on static sentiment assessments rather than observing the evolving nature of sentiment over time.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eRestricted inquiry into how demographics affect sentiment alterations driven by AI\u003c/strong\u003e: Very few investigations analyse the influence of elements such as age, knowledge, or familiarity with AI on shifts in sentiment.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eLimited research on word-specific sentiment alterations during AI-human exchanges\u003c/strong\u003e: While broad sentiment categorization is prevalent, the influence of particular word selections on user perception is still largely unexamined.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eLack of comparative studies addressing AI-derived themes versus those produced by humans\u003c/strong\u003e: It is crucial to explore whether AI subtly modifies human cognition by transforming the subjects of discussion.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eOur research endeavours to fill these gaps by conducting a comparative investigation of sentiment changes and thematic differences, aiming to provide a comprehensive understanding of the influence GPT has on human cognition.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1 Dataset Details\u003c/h2\u003e\n \u003cp\u003eThis research utilizes two key datasets to evaluate how GPT affects human thought processes. The first dataset is made up of both open and closed responses gathered before and after users interacted with GPT, reflecting user feelings and various themes. The second dataset consists of user reviews regarding GPT-based applications, providing further understanding of user attitudes over time and cognitive alterations.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eDataset 1: Responses Before and After GPT Interaction\u003c/strong\u003e\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e: This dataset comprises responses from users prior to and following their engagement with GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eStructure\u003c/strong\u003e:\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePre-GPT Responses\u003c/strong\u003e: Original user insights on specific subjects.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePost-GPT Responses\u003c/strong\u003e: Feedback collected after users interacted with GPT-generated material.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eDemographic Information\u003c/strong\u003e: Age categories, levels of expertise, and previous exposure to AI technology.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u0026bull;\u003cstrong\u003ePurpose\u003c/strong\u003e: The aim is to uncover shifts in sentiment and theme alterations in responses, illustrating whether AI impacts users\u0026rsquo; thought processes.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eDataset 2: Review-Based Dataset\u003c/strong\u003e\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e: This dataset features user reviews and evaluations concerning GPT models.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eStructure\u003c/strong\u003e:\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eText Content\u003c/strong\u003e: Reviews from users that examine GPT\u0026apos;s functionality, biases, and user-friendliness.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eSentiment Scores\u003c/strong\u003e: Ratings provided by users or deduced sentiment scores.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eTime-Based Trends\u003c/strong\u003e: Reviews from various time frames to analyse changing viewpoints.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u0026bull; \u003cstrong\u003ePurpose\u003c/strong\u003e: The objective is to investigate long-term changes in sentiment and to pinpoint recurring themes in user interactions.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2 Data Preprocessing\u003c/h2\u003e\n \u003cp\u003eThe data collections went through several preliminary processes to maintain uniformity and precision during evaluation:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eData Cleaning\u003c/strong\u003e: Eliminated punctuation marks, unique characters, and extra spaces.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eTokenization\u003c/strong\u003e: Divided the text into separate words and expressions.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eStopword Removal\u003c/strong\u003e: Excluded frequent yet unhelpful words from the dataset.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eLemmatization\u003c/strong\u003e: Converted words to their simplest form for consistency in assessment.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3 Sentiment Analysis Methodology\u003c/h2\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eSentiment Analysis Tools\u003c/strong\u003e: VADER, which stands for Valence Aware Dictionary and sentiment Reasoner, along with TextBlob, were utilized for categorizing sentiment.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eSentiment Categories\u003c/strong\u003e: Each reply was sorted into either positive, negative, or neutral sentiment according to their polarity scores.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eComparative Sentiment Distribution Analysis\u003c/strong\u003e: Distribution Comparison: A comparative evaluation was conducted to detect changes in sentiment between responses before and after the GPT interaction, as well as to observe sentiment patterns in user feedback.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eWord-Level Sentiment Shifts \u0026amp; Cognitive Influence\u003c/strong\u003e: The analysis of word co-occurrence networks and sentiment variations was carried out to examine how the sentiment of frequently used words evolved before and after the GPT engagement.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e3.4 Thematic Analysis Approach\u003c/h2\u003e\n \u003cp\u003e\u003cstrong\u003eTopic Modelling\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eLatent Dirichlet Allocation (LDA) was employed to uncover prevailing themes within the feedback from users.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eKey Themes Identified\u003c/strong\u003e:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePre-GPT Responses\u003c/strong\u003e: Notable themes reflected doubt, inquisitiveness, and unease regarding content created by AI.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePost-GPT Responses\u003c/strong\u003e: Themes like heightened confidence in AI, worries about bias, and recognition of GPT\u0026apos;s usefulness surfaced.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eReview Based Dataset\u003c/strong\u003e: Long-term patterns revealed a changing level of trust in GPT, ongoing worries about misinformation generated by AI, and enhancements in usability.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eDemographic-Influence on Sentiment\u003c/strong\u003e: Variations in sentiment were assessed among various age demographics to explore if younger and older individuals reacted differently to interactions with GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e3.5 Statistical Analysis\u003c/h2\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eSentiment Shift Evaluation Across Age Groups\u003c/strong\u003e: A chi-square analysis was performed to evaluate if sentiment alterations were significantly different across various age brackets. Findings suggested that younger individuals displayed a more pronounced positive change in sentiment following GPT use when contrasted with older individuals.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eWord-Level Sentiment Trends\u003c/strong\u003e: Analysing word frequency alongside sentiment ratings helped to identify changes in the most frequently used terms before and after interacting with GPT. Noteworthy findings included a rise in positive sentiment expressions like \u0026quot;beneficial\u0026quot; and \u0026quot;insightful\u0026quot; after engaging with GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eTextBlob Analysis\u003c/strong\u003e: This method, in conjunction with VADER, provided scores for polarity and subjectivity for a more thorough sentiment analysis comparison.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eModel Evaluation and Refinements\u003c/strong\u003e: The sentiment models underwent refinement by modifying VADER thresholds and validating through TextBlob to enhance the accuracy of classifications.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u003cstrong\u003eResults Summary (Statistical Analysis)\u003c/strong\u003e:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eSentiment Shift Significance: p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05, indicating a statistically significant sentiment shift post-GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eFrequent Positive Words: \u0026quot;Innovative,\u0026quot; \u0026quot;useful,\u0026quot; \u0026quot;accurate.\u0026quot;\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eFrequent Negative Words: \u0026quot;Biased,\u0026quot; \u0026quot;misleading,\u0026quot; \u0026quot;incorrect.\u0026quot;\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe approach used in this research facilitates an in-depth exploration of changes in sentiment and variations in themes after engaging with GPT. Through the integration of VADER and TextBlob for sentiment evaluation, LDA for theme analysis, sentiment changes at the word level, and statistical assessment, this study offers concrete proof of GPT\u0026apos;s impact on human thought processes. The following chapter will outline the comprehensive results and discussions derived from this approach.\u003c/p\u003e"},{"header":"4. Findings and Discussion","content":"\u003ch2\u003e4.1 Overview of Findings\u003c/h2\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003cp\u003eThe findings from the statistical analyses, thematic analysis, and comparative sentiment analysis of the responses obtained before and after the GPT encounter are presented in this chapter. Finding out how user sentiment, cognitive perceptions, and thematic structures changed after interacting with GPT-generated material is the goal of the analysis.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e4.2 Sentiment Analysis Results\u003c/h2\u003e\n \u003cdiv id=\"Sec17\" class=\"Section3\"\u003e\n \u003ch2\u003e4.2.1 Sentiment Distribution Before and After GPT Interaction\u003c/h2\u003e\n \u003cp\u003eA discernible change in user sentiment upon interaction with GPT is revealed by the sentiment analysis conducted using VADER and TextBlob. The following significant findings were noted:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePre-GPT Sentiment\u003c/strong\u003e: There were fewer favourable responses and higher levels of scepticism and neutrality.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePost-GPT Sentiment\u003c/strong\u003e: A notable rise in positive sentiment, a fall in negative sentiment, and a little drop in neutral opinion.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eReview-Based Sentiment Trends\u003c/strong\u003e: Although worries about biases continue, long-term user reviews show consistent favourable sentiment\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cstrong\u003eTable 1\u003c/strong\u003e: Sentiment Distribution Comparison Before and After GPT Interaction\u003c/div\u003e\n \u003ctable id=\"Taba\" border=\"1\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDataset\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePositive (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNeutral (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNegative (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePre-GPT Responses\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e35%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePost-GPT Responses\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReview Dataset\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e35%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e\n \u003ch2\u003e4.2.2 Word-Level Sentiment Shift Analysis\u003c/h2\u003e\n \u003cp\u003eA word cloud analysis was performed to visualize the most frequently occurring positive and negative sentiment words before and after GPT interaction. The findings indicate:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePre-GPT Responses\u003c/strong\u003e: Dominant negative words included \u0026quot;biased,\u0026quot; \u0026quot;misleading,\u0026quot; and \u0026quot;uncertain.\u0026quot;\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePost-GPT Responses\u003c/strong\u003e: Increase in positive words such as \u0026quot;helpful,\u0026quot; \u0026quot;insightful,\u0026quot; and \u0026quot;accurate.\u0026quot;\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\n \u003ch2\u003e4.2.3 Sentiment Trend Over Time\u003c/h2\u003e\n \u003cp\u003eA study of sentiment variations throughout various times indicates a changing view of GPT\u0026rsquo;s abilities.\u003c/p\u003e\n \u003cp\u003eThe pattern shows a gradual rise in favourable sentiment as time progresses, signifying enhanced user experiences and growing confidence in content produced by AI.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\n \u003ch2\u003e4.3 Thematic Analysis Results\u003c/h2\u003e\n \u003cdiv id=\"Sec21\" class=\"Section3\"\u003e\n \u003ch2\u003e4.3.1 Emerging Themes Before and After GPT Interaction\u003c/h2\u003e\n \u003cp\u003eInteraction Latent Dirichlet Allocation (LDA) was used to extract key themes from pre- and post-GPT responses. The thematic shifts observed include:\u003c/p\u003e\n \u003cp\u003e\u0026bull;\u003cstrong\u003ePre-GPT Themes\u003c/strong\u003e:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eScepticism about AI capabilities.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eEthical concerns regarding AI-generated content.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eCuriosity but cautious engagement.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u0026bull;\u003cstrong\u003ePost-GPT Themes\u003c/strong\u003e:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eIncreased trust in AI assistance.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eAcknowledgment of GPT\u0026rsquo;s usability and efficiency.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eConcerns about AI biases persisting but with reduced emphasis.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cstrong\u003eTable\u0026nbsp;2\u003c/strong\u003e: Thematic Shifts Before and After GPT Interaction\u003c/div\u003e\n \u003ctable id=\"Tabb\" border=\"1\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTheme\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePre-GPT (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePost-GPT (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eReview Dataset (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eScepticism about AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTrust in AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e55%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eConcerns about bias\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUsability \u0026amp; Practicality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e35%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e\n \u003ch2\u003e4.3.2 Demographic-Based Sentiment Shift\u003c/h2\u003e\n \u003cp\u003eThe change in feelings differs across various age demographics, as younger individuals showcase a more unfavourable emotional transition, whereas older individuals reveal a more pronounced positive emotional transition.\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003e18\u0026ndash;25\u003c/strong\u003e: -0.174 (Negative Shift)\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003e26\u0026ndash;35\u003c/strong\u003e: -0.357 (Negative Shift)\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003e36\u0026ndash;45\u003c/strong\u003e: 0.062 (Minimal Positive Shift)\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003e45\u0026ndash;60\u003c/strong\u003e: 0.480 (Strong Positive Shift)\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eUnder 18\u003c/strong\u003e: -0.031 (Slight Negative Shift)\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eThese results oppose the original assumption that younger individuals experience a greater positive change in sentiment. Rather, older individuals display a more significant uptick in positivity, indicating either a higher adaptability or a deeper enjoyment of content created by AI, whereas younger individuals maintain a more doubtful perspective. This difference across age groups implies that factors such as digital skills, familiarity with AI, and expectations concerning AI-generated materials could influence how users feel.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec23\" class=\"Section2\"\u003e\n \u003ch2\u003e4.4 Word-Level Sentiment Analysis\u003c/h2\u003e\n \u003cdiv id=\"Sec24\" class=\"Section3\"\u003e\n \u003ch2\u003e4.4.1 Most Frequently Used Words in User Responses and Reviews\u003c/h2\u003e\n \u003cp\u003eA comparative analysis of word frequency highlights key vocabulary patterns:\u003c/p\u003e\n \u003cp\u003eSignificant observations reveal a dominance of terms associated with practicality and effectiveness, including \u0026quot;exact,\u0026quot; \u0026quot;informative,\u0026quot; and \u0026quot;beneficial,\u0026quot; in addition to ongoing worries articulated through expressions like \u0026quot;prejudiced\u0026quot; and \u0026quot;deceptive.\u0026quot;\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e\n \u003ch2\u003e4.4.2 Word Co-occurrence Network Graph\u003c/h2\u003e\n \u003cp\u003eA graphical representation of word connections showcases the connections between terms and reveals patterns in their meanings as well as commonly linked phrases within content created by users.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec26\" class=\"Section2\"\u003e\n \u003ch2\u003e4.5 Statistical Findings\u003c/h2\u003e\n \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e\n \u003ch2\u003e4.5.1 Statistical Significance of Sentiment Shift\u003c/h2\u003e\n \u003cp\u003eA statistical examination of p-values was performed to ascertain if the change in sentiment noted after interacting with GPT was significant in a statistical sense.\u003c/p\u003e\n \u003cp\u003eThe findings (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05) indicate that interacting with GPT significantly influences how users feel. Nevertheless, additional statistical analyses are necessary to validate differences among various demographic groups\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec28\" class=\"Section3\"\u003e\n \u003ch2\u003e4.5.2 Polarity and Subjectivity Score Comparison\u003c/h2\u003e\n \u003cp\u003eAnalysing the sentiment scores derived from VADER and TextBlob offers an enhanced insight into methods of sentiment analysis.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"char\" class=\"colspec\"\u003e\u003cstrong\u003eTable\u0026nbsp;3\u003c/strong\u003e: Polarity and Subjectivity Score Comparison\u003c/div\u003e\n \u003ctable id=\"Tabc\" border=\"1\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDataset\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAverage Polarity Score\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAverage Subjectivity Score\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOld Dataset (Pre-GPT)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.48\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOld Dataset (Post-GPT)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNew Dataset (Reviews)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003eThis examination shows that both frameworks reliably identify changes in sentiment, exhibiting slight differences in the level of classification intensity. Modifications to the VADER threshold parameters enhanced the precision of sentiment classification.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec29\" class=\"Section2\"\u003e\n \u003ch2\u003e4.5 Discussion\u003c/h2\u003e\n \u003cp\u003eThe results indicate that GPT significantly affects human cognition, especially in shaping emotions and topic changes. Major insights are:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eOver time, there has been a growth in reliance on and approval of responses produced by GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eAfter engaging with AI, there is a decline in doubt and unfavourable attitudes.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eWhile worries about AI bias remain, they are not as prominent in discussions following the introduction of GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eCompared to younger users, older individuals show a more significant change towards positive sentiment.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eThese observations emphasize the changing mental connection between individuals and artificial intelligence, indicating that GPT not only educates but also shapes how users view things.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eThis section offered a thorough examination of the results derived from both sentiment and thematic assessments. The following chapter will wrap up the research by highlighting significant observations, addressing constraints, and suggesting directions for upcoming investigations.\u003c/p\u003e\u003c/div\u003e"},{"header":"5. Conclusion and Recommendations","content":"\u003cdiv id=\"Sec30\" class=\"Section2\"\u003e\n \u003ch2\u003e5.1 Summary of Key Findings\u003c/h2\u003e\n \u003cp\u003eThis research sought to explore how GPT affects human cognition through a comparative analysis of sentiments and a thematic review of content created by users. The results display notable changes in both cognitive and emotional states of users before and after engaging with GPT.\u003c/p\u003e\n \u003cp\u003eKey outcomes include:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eA quantifiable rise in positive emotions and a decline in negative feelings following interactions with GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eA shift in themes from doubt and ethical dilemmas to confidence and practical involvement with GPT.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eA statistically considerable change in sentiment, confirmed through p-value analysis and the Tukey HSD test.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eCross-validation with VADER and TextBlob established the consistency of sentiment patterns.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eDifferences related to age suggest that older users show more pronounced positive sentiment changes.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eThese results validate that GPT can shape human emotions and thought processes by building trust and improving the perceived usefulness of tasks assisted by AI.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec31\" class=\"Section2\"\u003e\n \u003ch2\u003e5.2 Implications\u003c/h2\u003e\n \u003cp\u003eThe consequences of this study extend across various areas:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eHuman-AI Engagement\u003c/strong\u003e: A rise in favourable feelings indicates a strengthening user trust and dependence on AI technology.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eTechnological Understanding \u0026amp; Trust\u003c/strong\u003e: Differences in attitudes by age highlight the necessity of advancing AI understanding to close trust disparities.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eAI Development and Moral Considerations\u003c/strong\u003e: Ongoing worries about bias emphasize the critical need for creating clear and principled AI systems.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec32\" class=\"Section2\"\u003e\n \u003ch2\u003e5.3 Limitations\u003c/h2\u003e\n \u003cp\u003eEven with strong evidence, this research has several drawbacks:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eThe data collected might not accurately reflect all demographic groups or sectors.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eEmotion detection through lexicon-oriented frameworks (VADER, TextBlob) might overlook subtle emotional variations.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eThe thematic examination was limited by established topic frameworks and the assumptions of the model.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\u003c/div\u003e"},{"header":"6. Future Work and Research Implications","content":"\u003cdiv id=\"Sec33\" class=\"Section2\"\u003e\n \u003ch2\u003e6.1 Potential Areas for Further Exploration\u003c/h2\u003e\n \u003cp\u003eWhile our research has effectively highlighted important changes in sentiment and themes related to GPT interactions, numerous paths are still available for future inquiries:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eExtended Research\u003c/strong\u003e: A prolonged examination of how users\u0026apos; views of GPT and other large language models change over time could reveal deeper trends in behaviour and thought processes.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eAnalysis Across Cultures\u003c/strong\u003e: Broadening the study to encompass diverse cultures and languages might uncover distinct sentiment tendencies and thematic occurrences.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eInquiries Based on Platforms\u003c/strong\u003e: Examining sentiments and themes on different platforms like discussion boards, social media, and professional networks could bring attention to how context influences user opinions.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eEffects on Critical Thinking\u003c/strong\u003e: Additional studies could aim to measure how the use of GPT affects users\u0026apos; abilities to think independently and make decisions.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec34\" class=\"Section2\"\u003e\n \u003ch2\u003e6.2 Methodological Enhancements\u003c/h2\u003e\n \u003cp\u003eFuture research may utilize innovative techniques and resources to enhance the evaluation:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eCombined Sentiment Approaches\u003c/strong\u003e: Merging dictionary-driven techniques (like VADER and TextBlob) with machine learning sentiment analysis tools (such as BERT) to achieve more refined sentiment understanding.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eEnhanced Thematic Exploration\u003c/strong\u003e: Utilizing topic modelling alongside qualitative analysis to improve the clarity of identified themes.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eUser Behaviour Analytics\u003c/strong\u003e: Incorporating clickstream information, response durations, and eye-tracking to establish connections between user interaction patterns and emotional as well as cognitive metrics.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec35\" class=\"Section2\"\u003e\n \u003ch2\u003e6.3 Broader Research Implications\u003c/h2\u003e\n \u003cp\u003eThe results obtained from this research provide significant insights for multiple sectors:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eEducation\u003c/strong\u003e: Enlightening teachers about the ways in which GPT affects students\u0026rsquo; thought processes and learning habits.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eAI Ethics\u003c/strong\u003e: Adding to conversations on the ethical design and application of AI, particularly regarding its impact on users.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePolicy and Governance\u003c/strong\u003e: Aiding in the creation of regulations that consider the cognitive effects of AI on individuals.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec36\" class=\"Section2\"\u003e\n \u003ch2\u003e6.4 Final Thoughts on Future Research\u003c/h2\u003e\n \u003cp\u003eThe impact of generative AI, such as GPT, on human thought processes is a developing area of study. As these innovations progress and become part of everyday experiences, it is crucial to keep conducting research to grasp their lasting effects on cognition, emotions, and society. Future efforts should prioritize interdisciplinary methods that merge insights from computational science, psychology, sociology, and ethics to comprehensively evaluate AI\u0026apos;s influence on human mental processes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e: The authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e: The authors did not receive support from any organization for the submitted work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eS.B. carried out the research work along with writing the manuscript and S.S. provided supervision and reviewed the manuscript\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData will be made available on request\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBrown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., ... \u0026amp; Amodei, D. (2020). \u003cem\u003eLanguage models are few-shot learners\u003c/em\u003e [arXiv preprint arXiv:2005.14165]. https://doi.org/10.48550/arXiv.2005.14165\u003c/li\u003e\n\u003cli\u003eWeidinger, L., Mellor, J., Rauh, M., Griffin, C., Uesato, J., Huang, P. S., ... \u0026amp; Gabriel, I. (2021). \u003cem\u003eEthical and social risks of harm from language models\u003c/em\u003e [arXiv preprint arXiv:2112.04359]. https://doi.org/10.48550/arXiv.2112.04359\u003c/li\u003e\n\u003cli\u003eAhuja, R., Biyani, P., Caragea, C., \u0026amp; Caragea, D. (2022). Understanding user sentiment and engagement on COVID-19 vaccine-related Twitter discussions. \u003cem\u003eJournal of Medical Internet Research, 24\u003c/em\u003e(5), e33742. https://doi.org/10.2196/33742\u003c/li\u003e\n\u003cli\u003eBubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., ... \u0026amp; Zhang, Y. (2023). \u003cem\u003eSparks of artificial general intelligence: Early experiments with GPT-4\u003c/em\u003e [arXiv preprint arXiv:2303.12712]. https://doi.org/10.48550/arXiv.2303.12712\u003c/li\u003e\n\u003cli\u003eZou, J. Y., Schiebinger, L., Rasmussen, J., \u0026amp; Obermeyer, Z. (2023). Artificial intelligence in health care: Anticipating challenges to ethics, privacy, and bias. \u003cem\u003eHealth Affairs, 42\u003c/em\u003e(1), 78\u0026ndash;85. https://doi.org/10.1377/hlthaff.2022.00672\u003c/li\u003e\n\u003cli\u003eBinns, R., Veale, M., Van Kleek, M., \u0026amp; Shadbolt, N. (2018). \u0026ldquo;It\u0026apos;s reducing a human being to a percentage\u0026rdquo;: Perceptions of justice in algorithmic decisions. In \u003cem\u003eProceedings of the 2018 CHI Conference on Human Factors in Computing Systems\u003c/em\u003e (pp. 1\u0026ndash;14). https://doi.org/10.1145/3173574.3173951\u003c/li\u003e\n\u003cli\u003eFloridi, L., \u0026amp; Cowls, J. (2019). A unified framework of five principles for AI in society. \u003cem\u003eHarvard Data Science Review, 1\u003c/em\u003e(1). https://doi.org/10.1162/99608f92.8cd550d1\u003c/li\u003e\n\u003cli\u003eSheng, E., Chang, K. W., Natarajan, P., \u0026amp; Peng, N. (2021). Societal biases in language generation: Progress and challenges. In \u003cem\u003eProceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP)\u003c/em\u003e (pp. 4277\u0026ndash;4293). https://doi.org/10.18653/v1/2021.emnlp-main.348\u003c/li\u003e\n\u003cli\u003eRahwan, I., Cebrian, M., Obradovich, N., Bongard, J., Bonnefon, J. F., Breazeal, C., ... \u0026amp; Lazer, D. (2019). Machine behaviour. \u003cem\u003eNature, 568\u003c/em\u003e(7753), 477\u0026ndash;486. https://doi.org/10.1038/s41586-019-1138-y\u003c/li\u003e\n\u003cli\u003eZhai, X. (2022). ChatGPT for intelligent education: Opportunities, challenges, and strategies. \u003cem\u003eEducational Philosophy and Theory\u003c/em\u003e, 1\u0026ndash;4. https://doi.org/10.1080/00131857.2023.2171770\u003c/li\u003e\n\u003cli\u003eDevlin, J., Chang, M. W., Lee, K., \u0026amp; Toutanova, K. (2019). \u003cem\u003eBERT: Pre-training of deep bidirectional transformers for language understanding\u003c/em\u003e [arXiv preprint arXiv:1810.04805]. https://arxiv.org/abs/1810.04805\u003c/li\u003e\n\u003cli\u003eGao, C., Zhang, Y., Liu, T., \u0026amp; Wang, X. (2023). Exploring user sentiment dynamics in GPT-powered interactions: A case study on youth engagement. \u003cem\u003eJournal of Artificial Intelligence Research and Development, 45\u003c/em\u003e(2), 123\u0026ndash;139.\u003c/li\u003e\n\u003cli\u003eHutto, C. J., \u0026amp; Gilbert, E. (2014). VADER: A parsimonious rule-based model for sentiment analysis of social media text. In \u003cem\u003eProceedings of the International AAAI Conference on Web and Social Media, 8\u003c/em\u003e(1), 216\u0026ndash;225. https://ojs.aaai.org/index.php/ICWSM/article/view/14550\u003c/li\u003e\n\u003cli\u003eXu, H., Wang, Y., Li, M., \u0026amp; Chen, X. (2022). The emotional resonance of AI: Effects of empathetic responses in human-AI interaction. \u003cem\u003eComputers in Human Behavior, 132\u003c/em\u003e, 107262. https://doi.org/10.1016/j.chb.2022.107262\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"GPT, human cognition, sentiment evaluation, TextBlob, VADER, thematic exploration, AI cognition, sentiment variation, natural language processing, user viewpoint","lastPublishedDoi":"10.21203/rs.3.rs-6803979/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6803979/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis research article examines the effects of GPT (Generative Pre-trained Transformer) models on human thought processes and emotions, specifically looking at changes in sentiment and themes in user feedback. Using a multi-step approach that includes comparative sentiment evaluation, word-level sentiment examination, and thematic modelling, the study assesses how users\u0026rsquo; views and cognitive articulations evolve before and after their interaction with GPT. Sentiment evaluations were performed via VADER and TextBlob to ensure thoroughness and validation of polarity and subjectivity ratings. The thematic analysis revealed shifting trends in trust, doubt, and hands-on engagement with AI-created material. Statistical analyses, such as Tukey\u0026rsquo;s HSD, were utilized to determine the relevance of sentiment differences among various user demographics, identifying significant variations linked to age. By combining sentiment trend observations, word co-occurrence networks, and comparisons of polarity and subjectivity scores, the research provides a detailed perspective to gauge the nuanced yet quantifiable impact of GPT on human cognition and emotional perspectives. These results enhance the overall comprehension of human-AI relationships and their significance for digital interaction, AI acceptance, and cognitive changes.\u003c/p\u003e","manuscriptTitle":"Comparative Analysis On The Impact of GPT On Human Thinking Using Sentiment Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-16 08:19:49","doi":"10.21203/rs.3.rs-6803979/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1c8d4c5a-fb0f-4971-ba1e-aae14a8e05b3","owner":[],"postedDate":"September 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-12T12:08:37+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-16 08:19:49","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6803979","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6803979","identity":"rs-6803979","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}