Folk Psychological Attributions of Consciousness to Large Language Models

Colombatto, C., & Fleming, S. M.

(2023, November 22). PsyArXiv

https://doi.org/10.31234/osf.io/5cnrv

Abstract

Technological advances raise new puzzles and challenges for cognitive science and the study of how humans think about and interact with artificial intelligence (AI). For example, the advent of Large Language Models and their human-like linguistic abilities has raised substantial debate regarding whether or not AI could be conscious. Here we consider the question of whether AI could have subjective experiences such as feelings and sensations (“phenomenological consciousness”). While experts from many fieldshave weighed in on this issue in academic and public discourse, it remains unknown how the general population attributes phenomenology to AI. We surveyed a sample of US residents (N=300) and found that a majority of participants were willing to attribute phenomenological consciousness to LLMs. These attributions were robust, as they predicted attributions of mental states typically associated with phenomenology –but also flexible, as they were sensitive to individual differences such as usage frequency. Overall, these results show how folk intuitions about AI consciousness can diverge from expert intuitions –with important implications for the legal and ethical status of AI.

The preprint is here.

My summary:

The results of the study show that people are generally more likely to attribute consciousness to LLMs than to other non-human entities, such as animals, plants, and robots. However, the level of consciousness attributed to LLMs is still relatively low, with most participants rating them as less conscious than humans. The authors argue that these findings reflect the influence of folk psychology, which is the tendency to explain the behavior of others in terms of mental states.

The authors also found that people's attributions of consciousness to LLMs were influenced by their beliefs about the nature of consciousness and their familiarity with LLMs. Participants who were more familiar with LLMs were more likely to attribute consciousness to them, and participants who believed that consciousness is a product of complex computation were also more likely to attribute consciousness to LLMs.

Overall, the study suggests that people are generally open to the possibility that LLMs may be conscious, but they also recognize that LLMs are not as conscious as humans. These findings have implications for the development and use of LLMs, as they suggest that people may be more willing to trust and interact with LLMs that they believe are conscious.

Deconstructing Moral Character Judgments

Hartman, R., Blakey, W., & Gray, K.

Current Opinions in Psychology

Abstract

People often make judgments of others' moral character-an inferred moral essence that presumably predicts moral behavior. We first define moral character and explore why people make character judgments before outlining three key elements that drive character judgments: behavior (good vs. bad, norm violations, and deliberation), mind (intentions, explanations, capacities), and identity (appearance, social groups, and warmth). We also provide a taxonomy of moral character that goes beyond simply good vs. evil. Drawing from the Theory of Dyadic Morality, we outline a two-dimensional triangular space of character judgments (valence and strength/agency), with three key corners-heroes, villains, and victims. Varieties of perceived moral character include saints and demons, strivers/sinners and opportunists, the non-moral, virtuous and culpable victims, and pure victims.

Conclusion 

It seems obvious that people make summary judgments of others’ moral character, but less obvious is how exactly that make those judgments. We suggest that people rely upon behavior, identity, and perceived mind when inferring the moral essence of others. We acknowledge that this list is certainly incomplete and will be expanded with future research. One key area of expansion explored here is the importance of perceived strength/agency in character judgments, which helps provide a taxonomy of character types. Whatever the exact varieties and drivers of moral character judgments, these judgments are clearly an important foundation of social life.

A pdf is here.

Doctors are seen as Godlike: Moral typecasting in medicine

A. Goranson, P. Sheeran, J. Katz, & K. Gray
Social Science & Medicine
Available online 25 May 2020, 113008

Abstract

Doctors are generally thought of as very intelligent and capable. This perception has upsides—doctors are afforded respect and esteem—but it may also have downsides, such as neglecting the mental and physical health of physicians. Two studies examine how Americans “typecast” doctors as Godlike “thinkers” who help others, rather than as vulnerable “feelers” who might themselves need help.

Highlights

• Americans view doctors as godlike and invulnerable.

• Doctors are seen as more agentic than other working professionals.

• Doctors are seen as able to ignore mental and physical health problems.

• Moral typecasting in medicine leads people to neglect doctors' suffering.

From the Discussion

Indeed, doctors are seen as equal to God in their capacity to think, exert self-control, remember details, and plan for the future (see Figure 1). Past work reveals that people typecast those who help others both high in agency and low in experience—which makes them invulnerable to injury and insult, and relatively incapable of suffering (K. Gray & Wegner, 2009). Our results confirm the existence of moral typecasting in medicine: compared to other working adults, people see doctors as less sensitive to pain, fear, embarrassment, and hunger (see Figure 2). We further find that these perceptions of super-human doctors extend outside of work and into global perceptions of physicians’ traits and abilities. This work adds to other research arguing that people do not want to acknowledge the feelings of healthcare providers, because this would make providers less capable of serving our health-related goals (Schroeder & Fishbach, 2015).

A pdf of the research is here.