HARNESSING GENERATIVE AI(GAI) TO MITIGATE INVESTOR'S BIAS

290 CHAPTER-26 HARNESSING GENERATIVE AI(GAI) TO MITIGATE INVESTOR’S BIAS Ashutosh Singh, Assistant Professor, Department of Commerce and Management GNIOT Institute of Professional Studies, Greater Noida, India Gbenga Festus Babarinde, Department of Banking and Finance, Modibbo Adama University, Yola, Nigeria Anil Kumar Gupta Assistant Professor, Department of Computer Science and Application Kamal Institute of Higher Education and Advance Technology affiliated to Guru Gobind Singh Indraprastha University, New Delhi, India Abstract This chapter examines the change-making role of Generative Artificial Intelligence (GAI) in reversing behavioral biases that impair rational investment choices. The behavior biases of confirmation bias, overconfidence, anchoring, loss aversion, and herding tend to cause market anomalies and poor financial decisions. GAI reverses these biases through real-time market information, historical patterns, and sophisticated analytics to offer unique, data-based insights. By virtue of such capabilities as sentiment analysis and portfolio diversification, GAI minimizes dependence on mental heuristics, encouraging decision-making that is well-informed. The chapter follows the history of AI in finance, from algorithmic trading to current robo-advisors, to illustrate the superiority of GAI in handling massive datasets. Nevertheless, such issues as algorithmic biases, datadriven hallucinations, and investor dependence challenge the importance of ethical design and effective regulation. By integrating behavioral finance and AI research, this chapter advocates a balanced approach that combines human intuition with GAI’s analytical prowess. It emphasizes the importance of financial literacy and regulatory frameworks to democratize GAI’s benefits, ensuring investors navigate complex markets with clarity and confidence. Keywords: Generative AI (GAI) Behavioral Bias, Decision Making, algorithum. Introduction In his latest writing, "Nexus: A Brief History of Information Networks from the Stone Age to AI," Yuval Noh Harari defied information not as factual data but also as shared belief, symbolic representations and cultural norms. The information can set a specific narrative, myths and intersubjective realities. During the digital epoch, an individual relies on digital information more than other sources. However, the author questioned the integrity of the AI models. AI may provide information that may be unreal[1]. The webspace has been changed. It is more commercialized and personalized. Advertisers are spying on every 291 product the customer is watching on their screen. Computer and smartphone screens are becoming a powerful medium for creating user perceptions. It has recently been found to be the strongest medium for polarising radical mentality. "The Filter Bubble," authored by Pariser, reveals how cyberspace controls our lives[2]. However, it is undoubtedly one of the most valuable tools for knowledge and information if used with discretion. The information our mind feeds, in combination with our personality traits and social status accompanying the environment, constructs behavioral biases. Such biases shape behavior and decision-making [3]–[5]. After the introduction of the Capital Assets Pricing Model (CAPM), Arbitrage Pricing Theory (ATP) and Efficient Market Hypothesis (EMH) theory, a discussion over the rationality of humans in making investment decisions took place[6][7]. Behavioral finance claims that the irrational behavior of investors is the cause of financial market anomalies[8], [9]. For a long time, behavioral biases have been treated as the cause of loss. It is a myth that behavioral biases always lead to losses. They may even lead to gains and sometimes a mix[10]. Behavioral biases build courage in investors to make riskier decisions and enhance decisions by reinforcing confidence in investment options. Behavioral biases minimize psychological chaos and help conclude distress [11]. For a long time, investors have dreamed of getting a computer program to help them design investment strategies. Computer programmers created several software programs that helped investors. However, they had limited availability to every investor. Algorithmic (Algo) trading was the first term introduced when Machine Learning (ML) technologies were introduced. Algo trading was featured as a computer program that automatically executes the trade without human intervention. Furthermore, AI technology gives faster and more accurate market information processing [12]. During the fourth industrial revolution, investors significantly used generative AI. The awareness about artificial intelligence has increased among investors. It also helps investors delve deep into investment strategies. AI also considers the information that human senses may ignore[13]. No doubt, generative AI is more intelligent than an experienced investment advisor. It searches previous datasets, theories of investment, documented strategies, and previous trends, and it also analyses the options. Anxiety, trust, and performance expectancy are significant factors in adopting AI or roboadvisors[14]. The introduction of technology in investment started in the 1950s. However, the model was not powerful enough to analyze the large datasets. It primarily focused on symbolic reasoning and algorithms. In 1980, more sophisticated algorithms were developed, but their computational efficiency was still limited. In 1990, quantitative finance, mathematical, and statistical models in finance were developed as a focal point of investment strategies. A growth in algorithms has been witnessed during this period. The years 2000 to 2010 witnessed the traction gained by the development of ML technologies. 2010-2020 is acknowledged as a 292 decade of big data and deep learning. AI started analyzing the data through the unstructured data available on social media sites and news. After 2020, the performance of AI investment advisors and their applications has intensified. AI applications generate suggestions more personal to the human touch[15]. There are some common pitfalls like overconfidence, anchoring on irrelevant information, and the list goes on. The brain of the investor uses shortcuts or heuristics for decision-making. They are helpful sometimes but can also guide the wrong path[16]. This chapter is an effort to introduce the readers to the application of generative AI in investment advice. This chapter also discusses several behavioral biases, and further, the chapter attempts to capture the viewpoints of prominent researchers in this area. Review of literatureInformation and Investor’s Bias Various information signals often influence the investor's decision-making process, leading to biased decisions. The narrative description of news can influence investor psychology; sometimes, it differs from the actual [17]. Due to several behavioral errors, investors make wrong investment decisions. When investors collectively follow the same trend, it results in a market anomaly. However, conventional theories, such as Arbitrage Pricing Theory[18], Efficient Market Hypothesis [19], Modern Portfolio Theory [20] and Capital Assets Pricing Model [21], state that investors are risk averse and rational in decision-making. On the contrary, the Prospect Theory[22]claims the above statements of behavior errors. Behavioral errors result from heuristics and biases [23], [24]. Failure or the decision-making success depends upon the investor's investment consumption pattern. Investors pay attention to the information that influences their beliefs, and decisions potentially align with the "naive" and "hardheaded" investor types[25]. Behavioral theorists listed several behavioral biases, such as overconfidence, herding, loss aversion, and Status Quo Bias [26]. Available information for the investor plays a vital role in making investment decisions. Information carves the investor's psychology, such as confidence and future orientation. An investor who gathers less information is much affected by cognitive bias. The financial satisfaction of investors minimizes the need for information search. It further leads to overconfidence in investors. Although the information is necessary for decision-making, it can instigate confirmation bias, information overload, framing effect, recency bias, and availability heuristics[27]. When making financial decisions during uncertainty with plenty of information available, investors take mental shortcuts that can lead to suboptimal choices[28]. While researching investment options, an investor becomes inclined toward local stocks compared to foreign ones. The information enhances the familiarity bias; investors make their judgments about choosing local stocks over foreign ones. The most common clarifications are information processing costs and uncertainty of foreign financial reporting [29]. The researchers suggest that investors get 293 biased in selecting investment options and acquire information that strengthens their beliefs. There is an intrinsic relationship between decision-making and confirmation bias[30]. Research in Kenya pointed out the information processing bias. Investors rely on opinions and external sources instead of their analysis. It resembles the herding behavior, but a herd or herding behavior is a collective action of investors[31]. Several kinds of literature portray investor biases as a villein that always harms the investor. Nevertheless, as the investor processes the signals of stock markets with the available information, the biases positively affect the decision-making [17].Investors do not flow with emotions; information enhances confidence and motivates reasonable decision-making; this is a positive aspect of biases[32]. Finally, the quality of the information matters in decision-making. Decisionmaking is a cognitive process that involves emotions. Complete information will enhance the understanding of the situation and lead to a firm decision. The role of information gets elevated during uncertainty like COVID-19. Publicly available information affected investor behavior and market trends [33]. After applying the Elaboration Likelihood Model (ELM), it inferred that due to the low ability and motivation of the investor, negative aspects of biases affect the psychology towards the market signals. In contrast, a motivated and able investor is affected by the positive aspect of biases [17].Biased behavior is not limited to retail investors or beginners, but the experts and planners are vulnerable to behavioral bias, which hinders their ability to provide the optimum investment solutions [28]. Generative AI and decision-making Generative AI is an Artificial Intelligence (AI) technology discovered in the 1960s; Chatbots were introduced in 2014. This technology helps develop highquality text, images and graphics. Furthermore, several AI products can analyze and comprehend text, images and video. Generative AI is used in several areas for decision-making. Chat GPT is shifting traditional medical practices towards a new dimension. Its potential exists within the access to the unlimited sources of information available on the internet. It can help generate detailed and coherent responses to medical queries that need immediate reaction. It can be helpful in accurately analyzing the symptoms. However, it generates potentially irrelevant results. The accuracy and reliability of GAI models are still in the experimental phase[34].Chat GPT cannot substitute an expert decision-maker, but after analyzing the circumstances from news and online resources, it gives several options for making decisions. It will be a challenge for online search engines[35].Researchers claim combining GAI with Human Intelligence can result in effective and creative decisions. GAI is capable of providing data-driven support and predictive analysis. Most of the time, human intelligence lacks datadriven skills. GAI optimizes the decision-making outcomes [36].Some scholars raised ethical concerns over using GAI in education and research. The primary concerns are legal and compliance issues, integrity, confidentiality, and plagiarism. Furthermore, biased or misleading information is also a concern[37]. 294 In finance, AI can generate data from large and real datasets. It can model complex systems, concepts, and patterns. It can predict future trends and anomalies From previous trends and patterns. Synchronizing information on the internet allows for real-time analysis and suggests decisions [38]. Financial markets face continuous uncertainty, efficiency and volatility. Such barriers hinder the entry of potential investors. Generative AI breaks down such barriers by offering customized and personalized investment strategies[39]. Integrating GAI with financial decisionmaking can optimize the process and add more personal insight[38]. GAI gives valuable insights to investors while analyzing the financial market, identifying frauds and risks, and interpreting unstructured data[36].AI Advisors utilizing Modern Portfolio Theory (MPT) and Machine Learning (ML) techniques can predict market movements and asset performance and maximize returns[40]. Confirmation Bias and GAI’s Counteractive Mechanisms Confirmation bias is when investors seek information supporting what they already think is a significant issue in investment decision-making. We at GAI address this by using many data sources, which include market trends, news, and social media feelings, to present a balanced point of view. Liu et al.[41]did a study that tested ChatGPT's role in gold investment decisions. They found that GAI reduces confirmation bias via multi-step zero-shot reasoning, challenging the investor's select info processing. Also, GAI, by its integration of historical data with present market signs, gets investors exposure to the other side of the argument, which, in turn speaking, expands their decision-making range. Also, GAI can look at unstructured data like social media posts, allowing it to put a finger on sentiment-based biases that play into confirmation tendencies. In a study done by Hao et al.[36]examined the collaboration between humans and GAI. It was found that GAI's data play a vital role in reducing the tendency to ignore, which is the opposite of confirming what we do not want to hear. They did a quasi-experimental study that reported that using GAI in prediction improves decision quality by 15%, which is against what we see with human-only decisions in very volatile markets. However, the authors warn that the poor design of prompts may put forward confirmation bias, which stresses the importance of good algorithm design. Overconfidence and GAI’s role in risk assessment. Overconfidence bias has investors overestimate their knowledge or predictive skills, which in turn causes them to take on too much risk. GAI, on the other hand, looks at trade history and presents evidence-based risk assessments. Hasan[42]looked at robo-advisors who use GAI and machine learning, which they found to be effective in identifying overconfident trading trends via portfolio concentration and trade frequency analysis. His study published in the Journal of Ecohumanism reported that GAI, behind diverse investment strategies, reduced portfolio volatility by 12% for overconfident investors. Also, by giving 295 transparent reasons for their recommendations, GAI increases investor trust, which in turn gets them to follow diverse investment plans. Also, we see that GAI, the gamification element in play here, improves financial literacy and reduces overconfidence. In a study by Ng et al.[43]of AI-based educational games that play out market scenarios, we find that the players who used these tools reported a 20% drop in overconfidence, which they had in their trading decisions. Also, in this study, which was published in the Journal of Computer Assisted Learning, they report that GAI puts the user in a 3D experience, which in turn allows investors to see the results of their overconfident actions in a safe setting. While we see the benefits of this approach, in a study by Bastani et al.[44], we are told that this may be too much of a good thing as they report that new investors who are very into the personal feedback from the GAI may become more confident which is an issue which requires that human input be included. Anchoring Bias and Real-Time GAI Interventions Anchoring bias causes investors to put great weight on past information, such as past stock performance, which may not be relevant in present conditions. GAI, which we have, is a step beyond this by which it changes recommendations in real time based on market signals. Southworth et al.[45]looked at the role of GAI in financial education, which we see that it uses real-time analysis of social media and news to determine market sentiment, which in turn reduces anchoring by as much as 18% as compared to the traditional advice which is doled out. Also, their study showed how GAI does a better job of getting investors to look at present market trends instead of past data, which may not be relevant. GAI models trained on past data may sometimes put out-of-date patterns, which requires continuous model retraining for relevance. Loss Aversion and GAI’s Behavioral Nudging Loss aversion, in which investors put more weight into avoiding losses than into making gains, causes them to make suboptimal portfolio choices. GAI corrects this via behavioral nudges and scenario-based simulations. Ng et al. [43]reported that AI-powered educational games that play out market dips improve decisionmaking. They have investors face loss situations in a simulated setting, which helps them overcome loss aversion. They found that people who underwent GAI simulation were 25% more likely to hold diverse portfolios through market downturns. By putting forth probability results, GAI re-frames losses within the context of long-term strategies, which in turn plays down emotional decisionmaking. Also, GAI's transparency in reporting on market changes plays a role in loss aversion by making volatility more understandable. Fatima and Chakraborty [14]did a study in India which reported that GAI's performance in the "why" of market movement issue raised investor confidence by 30%. Their study, which was published in IIMB Management Review, also revealed GAI's role in building 296 trust, which in turn plays a role in loss-averse behaviors. Also, as Chui et al.[46]put forth, GAI's persuasive communication style may sometimes put out too optimistic a picture, which may cause investors to have unrealistic expectations. Herding Behavior and GAI’s Individualized Insights Herding, where investors jump on trends without independent analysis, is a factor in market anomalies. GAI puts forth individualized insights based on investor and market information. Zhou et al.[47]looked at how AI puts out misinformation and what it does to herding; they found that GAI, which does real-time analysis of social media trends, helps investors out of herd mentalities. Their study reported a 20% drop in herd-driven trades among GAI users, resulting from the tech that brings contrarian strategies based on data to light. Also, we see that GAI's ability to identify sentiment bubbles in social media, which in turn improves its performance regarding herding. A study by Sai et al.[38]published in Expert Systems looked into GAI use in finance. They report that sentiment analysis, which in turn reduces herding by which they mean it pérdorms to break up irrational market exuberance. They also report that we see that GAI-driven robo advisors did better by 15%, which means they did better at preventing herd-driven losses during market bubbles. At the same time, however, Agrawal[48] reports that GAIs into which algorithms put too much stock of trendy sentiments as against fundamental analysis may amplify herding, which, in turn, he puts forth the case for algorithmic diversity. Implications Generative AI and Investor bias GAI is claimed to be free from biases and heuristics, while it is a limitation of human intelligence[36]. A key novelty of utilizing GAI is managing behavioral bias in financial decision-making. For some specific biases like confirmation and hindsight, AI technologies such as backpropagation within deep neural networks and deep reinforcement learning can help investors overcome them[28]. GAI can analyze historical data, present market signals, and statistical techniques to predict the future. Information adjoined with data analytics is more helpful than baseless news[42]. GAI presents a comprehensive viewpoint collected from several sources on the internet, analyses them on the base of algorithms, and further presents them to the investor[36]. AI may have several benefits in mitigating investor bias in the financial market, but experts still raise several concerns. The inherent training bias of the AI model may be amplified over time. Further, the structure of the prompts can influence the output and perpetuate existing bias [49]. GAI Mitigating Investor Bias GAI can be a valuable tool to mitigate behavioral bias. Investors trust the GAI because it can collect data from multiple sources and process and present it 297 understandably. GAI handles biases as it provides reliable answers to investors' queries. GAI has a remarkable knack for sifting through massive datasets, which include everything from historical market trends to real-time sentiment analysis and even unstructured data from social media. This capability allows it to tackle common behavioral biases like confirmation bias, overconfidence, and anchoring. For example, GAI can help reduce confirmation bias by pulling together various information sources, giving investors a more balanced view instead of just reinforcing what they already believe. A study by Liu et al. (2024) shows that tools like ChatGPT can lessen bias in gold investment decisions by offering datadriven insights that challenge the selective way investors process information. Likewise, GAI combats anchoring bias by adjusting recommendations based on the latest market signals, which helps investors avoid getting stuck on outdated reference points[41]. Hasan (2025) points out that robo-advisors that utilize GAI and machine learning can diversify portfolios to counteract overconfidence, as they analyze trading histories to spot risky patterns and suggest strategies to mitigate those risks[42]. GAI deals with confirmation bias progressively. At the same time, confirmation bias restricts an investor from searching and interpreting only favorable information. GAI integrates the data of the current news, trends, and signals of the stock market and analyses the historical data. Moreover, GAI counters the anchoring bias that confines investors to stick to outdated information; it uses realtime sentiment analysis on social media to give financial advice[45]. Although, GAI in the present stage cannot see the investor giving prompts. However, with the prompts, it analyses the personality and needs of the investor and provides a personalized result. AI technology can detect investors' overconfidence in their trading history. It can suggest diversifying the portfolio to minimize the risk[50].The AI technology is quite transparent. For beginners, the investor prefers to know the reason for market fluctuations. In the absence of available stock market experts, GAI answers this "Why" of investors. Awareness of the history of investment mistakes lets the investor avoid those mistakes. GAI plays a vital role in mitigating herding behavior. It prevents investors from following a specific trend by giving them real-time insight into signals. It has been witnessed that investors who unthinkingly follow social media advice get easily trapped into herding or herd behavior[47]. GAI in the financial literacy to the investor adds gravity in the gamification techniques of education and learning. Investor psychology practice to avoid lossaversion bias, improving decision-making over time. AI gamification has potential to change the investor behaviour for a long-term. Despite of the possibilities of mitigating the risk of biases it depends on the design of algorithms, that can introduce new challenges[43]. Ng et al. (2024) highlight how AI-driven educational games improve decision-making by simulating market scenarios, helping investors recognize and avoid loss-aversion bias[43]. These tools foster a 298 deeper understanding of market dynamics, empowering retail investors to make informed choices. By providing transparent explanations for recommendations, GAI also builds trust, addressing the psychological barriers to adopting AI advisors, such as anxiety or distrust[14]. GAI Accelerating Investor Bias While GAI models promise to reduce the risk of amplifying the existing biases or introducing new ones due to reliance on the historical data and limitations of the algorithm, societal prejudices and discrimination against certain groups can provide a biased response to GAI. Misalignment or divergence among the data source and reference, especially in the large dataset, does create a biased opinion[49].A GAI model generates factually incorrect information, which is called a hallucination. It may misinform the investor and lead to skewed decisionmaking. It can lead to overconfidence and optimism bias[51]. GAI tools are developed in the English language. It generates wrong results in languages other than English [45]. Accessibility and perceived reliability of GAI are increasing investors' addiction. Investors get addicted to searching for information on AI due to its personalization. Compared to traditional search engines, it provides the same results to every user, while GAI facilitates personalized results. Responses of the GAI are engaging, and investors receive instant gratification. Gradually, it creates a dependency on the application. Additionally, the human-like qualities of GAI attract the user. Empathy, natural language processing and emotional engagements enhance investor engagement with GAI applications [46].A habit of getting tailored responses mirrors addictive conduct. The beginners are prone to the GAI getting the confidence boost up inappropriately[44]. Regulatory and Practical Considerations The rapid adoption of GAI in financial markets necessitates robust regulatory frameworks to address its risks while harnessing its benefits. AI algorithms, including GAI, can amplify procyclicality and herding behavior if they adopt similar strategies across firms or rely heavily on social media sentiment. It underscores the need for regulatory guidelines that ensure algorithmic diversity and transparency. Given AI's evolving nature, the International Organization of Securities Commissions (IOSCO) suggests that supervisors focus on providing best practices rather than strict rules. For instance, regulators could mandate regular audits of GAI models to detect biases and ensure explainability [48]. Practically, the accessibility of GAI tools remains a barrier for retail investors, particularly in developing economies. High costs and technological infrastructure requirements limit GAI's reach, potentially widening the gap between institutional and individual investors. To democratize access, financial institutions should explore scalable, cost-effective GAI solutions, such as cloud-based robo-advisors. Additionally, policymakers should incentivize financial literacy programs that integrate GAI, ensuring investors understand its strengths and limitations [15].