The Impact of Algorithmic Predictions on User Behaviors in Online Investment Communities

Abstract

Algorithms are being increasingly adopted in online investment communities (OICs) to generate algorithmic predictions of stock performance for investors (i.e., OIC content consumers). Such predictions can supplement the analysis articles written by human analysts in OICs. Nevertheless, the introduction of algorithms may affect both human analysts and investors in OICs. On the one hand, human analysts are the main content providers who forecast stock performance through their content generation, i.e., stock analysis articles. Algorithmic predictions have functional similarity with human analysts in stock forecasts so that they can pose an identity threat to human analysts. On the other hand, investors are the primary content consumers in OICs. Investors can engage with stock analyses to voice dissenting views and resolve confusion, which enables them to obtain and internalize investment ideas effectively. Nevertheless, it remains unknown whether and how algorithmic predictions affect analysts’ and investors’ behaviors in OICs. Based on data from SeekingAlpha.com, and using the related theoretical lens as a framework, this thesis explores the impacts of algorithmic predictions on user behaviors in OICs.

This thesis first studies the impact of algorithmic predictions on analyst content generation. Drawing on identity control theory and coping theory, we theorize analysts’ coping responses under algorithmic identity threat as an algorithm reactance effect, which manifests in two forms, namely, algorithm differentiation and self-elevation. We further elaborate on the contingent role of algorithm performance and analyst experience in influencing this effect. Applying a regression discontinuity in time design, we found that analysts differentiated themselves from an algorithm in stock forecasts (i.e., algorithm differentiation). Specifically, analysts exhibit convergent differentiation with a slight deviation but in the same direction as the algorithm when the algorithm is accurate; by contrast, they exhibit divergent differentiation by opposing algorithmic predictions when the algorithm is inaccurate. Moreover, analysts demonstrated growing productivity (i.e., self-elevation), which is positively moderated by their experience. Our finding reveals a novel algorithm reactance effect in a human–algorithm interaction context and offers managerial implications for next-generation algorithm-empowered OICs.

This thesis also examines the impact of algorithmic predictions on investor engagement in OICs. Drawing upon information foraging theory, we theorize investors’ engagement adjustment strategy as a selective engagement effect, which manifests in two forms, namely, depth tension and portfolio diversification. We found that investors exhibit a tension on engagement depth to form confident opinion (i.e., depth tension). Specifically, novice investors or those facing divergent algorithmic and analyst forecasts intensify their engagement with stocks analyses. Conversely, experienced investors or those encountering aligned forecasts tend to mitigate their engagement. Moreover, investors increase engagement breadth across a broader set of stocks to diversify their investment choices (i.e., portfolio diversification). Our study uncovers a novel selective engagement effect in a human–algorithm interaction context and highlights the crucial contingencies influencing the tension within this effect.

In conclusion, this study examined the impact of the introduction of algorithmic predictions on user behavior in OICs, providing theoretical foundation and practical implications for literature in the field of human-algorithm/AI interaction. The research findings can assist the online investment community, analysts, investors and algorithm developers in decision-making. Our findings offer managerial implications for next-generation algorithm-empowered OICs.

Date of Award	25 Jun 2025
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Zhiya ZUO (Supervisor), Qiang Ye (External Supervisor), Weiquan WANG (Supervisor) & Juhee KWON (Co-supervisor)