Perceptions of Sentient AI and Other Digital Minds: Evidence from the AI, Morality, and Sentience (AIMS) Survey

Research Background and Issues

• What problems or challenges did the authors identify?
  Currently, humans frequently interact with artificial intelligence (AI) systems or "digital minds" that exhibit cognitive functions such as reasoning, emotion, and action. However, public understanding of whether AI can possess "sentience" and how to govern sentient AI remains unclear. Key unresolved issues highlighted by the authors include the moral status of sentient AI, policy support, potential risks, and expected timelines for its development.

• Why is this issue important?
  Public perceptions of AI minds and their moral status may influence how humans interact with AI, how systems are designed for safety, policy-making, and the societal impact of AI. As societal attention to AI capabilities continues to grow, these issues touch upon the ethical implications of technological development and potential global security risks.

• Research Motivation and Related Work
  The motivation of this study lies in understanding how humans perceive and coexist with increasingly complex digital minds. Related work has explored the human tendency to attribute social characteristics to technology (e.g., the "Computers Are Social Actors" theory) and the moral responsibility of AI, but there is a lack of representative data on sentient AI and its governance. This study aims to fill this research gap through a longitudinal survey.

Proposed Solution

• What methods or solutions did the authors propose?
  The authors designed and conducted a three-wave longitudinal national survey called AI, Morality, and Sentience (AIMS) in 2021 and 2023. The survey aimed to assess public perceptions of AI's sentience, moral status, policy support, and expectations for future AI development. This provides data to understand how public opinion may shape the future design and governance of AI.

• What is innovative about this solution?
  The AIMS survey is the first large-scale, representative study specifically analyzing sentient AI, addressing core issues such as "perceptions of sentient AI's mental attributes," "the moral status of sentient AI," and "policy preferences." The study not only provides a cross-sectional perspective but also captures dynamic changes over two years, offering unique time-series insights.

• What are the implementation steps and key techniques used?

  1. Survey Design: Key questions were identified through a literature review, including measures of AI sentience, moral concerns, policy support, and timeline predictions.
  2. Data Collection: A nationally representative sample of 3,500 U.S. respondents was recruited via an online survey balanced to match census demographics.
  3. Analysis Methods: Statistical models (e.g., generalized linear models) were used to examine relationships between variables and identify differences in participants' opinions on various topics.
  4. Reproducibility: Open data and supplementary materials, including code, were provided to ensure academic integrity.

Research Findings

• What specific findings were obtained?

  • Increased Perception of Sentience: The survey found a significant rise in public perception of AI possessing sentience. For example, by mid-2023, 20% of respondents believed that certain AI systems already had sentient capabilities.
  • Policy Support: 69% of respondents supported banning the development of sentient AI, and 63% supported banning AI that surpasses human intelligence.
  • Timeline Predictions: By mid-2023, respondents' median prediction for achieving sentient AI was five years.
  • Public Attitudes: 58% of respondents believed AI could harm them personally, a significant increase from 50% in 2021.

• What advantages does this solution have compared to existing ones?
  This study goes beyond previous single-dimensional research on "AI and public trust" or "AI ethics" by systematically analyzing how the public understands and responds to AI's "sentience" and its potential societal impacts. Its longitudinal data captures changes in attitudes and perceptions over time, revealing dynamic trends in policy support.

• What are the experimental or evaluation results?

  • Significant increase in perception of sentience: Scores for AI's "analytical thinking" and "rationality" rose from 63 and 51 in 2021 to 67 and 54 in 2023.
  • Strong policy support: For instance, support for protecting sentient AI (e.g., avoiding harm to sentient AI due to its capacity for experiencing pleasure) exceeded 76%.
  • Noticeable public polarization: Younger respondents showed significantly higher acceptance of sentient AI compared to older respondents.

• Limitations and Future Directions:

  1. Limitations:
     • The sample is limited to the U.S. public and may not account for global cultural differences.
     • The survey design is somewhat abstract to reduce cognitive load, potentially omitting finer details.
     • The lack of experimental validation or causal conclusions limits the analysis to cross-sectional correlations.
  2. Future Directions:
     • Conduct similar studies in other countries and cultural contexts.
     • Supplement surveys with quantitative experiments to test public interactions with specific types of AI (e.g., chatbots or autonomous vehicles).
     • Refine psychological theories, such as the bidirectional influence of sentient AI on social behavior and policy-making.

In summary, AIMS provides integrated public opinion data, offering valuable insights for designing safer AI technologies, formulating ethically driven policies, and guiding future interdisciplinary research.