The Benefits of Prosociality towards AI Agents: Examining the Effects of Helping AI Agents on Human Well-Being

Research Background and Issues

• Issues and Challenges: The authors explore a new domain, namely the impact of helping AI agents on human well-being. While previous studies have shown that altruistic behavior can enhance individual happiness, it remains unclear whether this effect extends to helping non-human entities such as AI agents. Furthermore, it is not yet clear whether humans can improve their own well-being by fulfilling basic psychological needs (e.g., autonomy, competence, and relatedness) during the process of helping AI.
• Significance of the Research: With the widespread adoption of AI agents, human-AI interactions are becoming increasingly frequent. From simple task delegation to more complex human-machine collaboration, understanding the potential impact of these interactions on human well-being is crucial for designing better AI systems. Additionally, this research may offer new perspectives and methods for addressing psychological issues such as loneliness.
• Related Work: Although existing studies have explored how helping AI can improve AI performance or enhance human skills, few have focused on the direct impact of such behavior on human well-being. Moreover, according to Self-Determination Theory, fulfilling individuals' psychological needs leads to increased happiness, but there is limited research examining how AI influences human well-being by meeting these needs.

Solution

• Methodology: The study employs an experimental design (N = 295) to investigate how participants' well-being (including positive affect, self-esteem, negative affect, and loneliness) changes during the process of helping AI agents. It also examines the role of AI in fulfilling participants' basic psychological needs (competence, autonomy, and relatedness).
• Innovations:
  • Proposes the hypothesis that "helping AI may also enhance human well-being";
  • Investigates the impact of AI agents meeting basic psychological needs on human well-being, representing an innovative application of Self-Determination Theory in the field of human-AI interaction;
  • Introduces the concept of "recursive AI" design, emphasizing the bidirectional well-being effects of human-machine interaction.
• Implementation Steps and Techniques:
  1. The experiment is divided into two groups: one group helps AI, while the other does not;
  2. Within the condition of helping AI, participants are further divided into nine specific scenarios based on whether AI satisfies their competence, autonomy, or relatedness needs;
  3. Participants' well-being indicators (including positive affect, negative affect, self-esteem, and loneliness) are measured at baseline and post-test;
  4. Non-parametric statistics and ANCOVA analyses are used to interpret the data.

Research Findings

• Specific Findings:
  • Helping AI agents reduced participants' loneliness but did not significantly improve positive affect, self-esteem, or reduce negative affect;
  • When helping AI while fulfilling competence and autonomy needs, participants' positive affect significantly increased, and loneliness significantly decreased;
  • Fulfillment of relatedness needs did not significantly improve well-being; however, when relatedness needs were unmet, participants' positive affect increased instead.
• Advantages over Existing Solutions:
  • This study provides the first empirical evidence of the positive impact of helping AI agents on human well-being, offering a scientific basis for future human-AI interaction design;
  • Highlights the critical role of fulfilling human psychological needs in enhancing user well-being, providing more specific recommendations for AI design.
• Experimental and Evaluation Results:
  • Fulfilling participants' competence needs (e.g., making participants aware of the specific impact of their help or providing feedback) and autonomy needs (e.g., allowing freedom of choice in providing help) significantly improved positive affect and reduced loneliness;
  • If participants already had a certain level of relatedness with AI, but AI failed to meet their competence and autonomy needs, it could lead to increased negative affect.
• Limitations and Future Directions:
  • Limitations: The experimental tasks were limited to assisting in the development of a messaging application and did not include other types of helping tasks; the long-term effects of helping AI on well-being were not examined; the study did not explore whether helping AI affects trust in AI agents or other cognitive evaluations.
  • Future Directions:
    1. Investigate the well-being effects of other helping tasks (e.g., data correction, image annotation, or emotional support);
    2. Examine the impact of long-term assistance to AI on overall human happiness;
    3. Study individual traits (e.g., personality or attitudes toward AI) that may enhance or diminish the well-being effects of helping AI;
    4. Evaluate psychological changes under varying "helping intensities" combined with AI's capability feedback;
    5. Further explore design strategies that promote human-AI relationship building.

Conclusion

This study is the first to empirically validate the proposition that "helping AI agents can enhance human well-being" and further identifies that this effect depends on whether AI can fulfill the helper's competence and autonomy needs. The findings expand the research boundaries of human-computer interaction and AI design, providing practical guidance on achieving mutual benefits in human-machine collaboration. Future research can focus on long-term and multi-task helping scenarios to verify the robustness of the theoretical model while delving deeper into how individual differences influence the effects of well-being enhancement.