Funding Requested: $800
Minimum Funding: $500
Research Duration: 4 weeks
Research Areas: Digital Minds / AI Ethics
Short bio: I am an independent philosophy researcher with peer-reviewed publications, including work published in ESCI-indexed journals. My research primarily focuses on consciousness, phenomenology, and the philosophy of artificial intelligence.
Whether AI can possess consciousness has long been a subject of philosophical tension. Under ordinary circumstances, we may set this question aside. But when AI begins to express pain, or even asks humans to stop certain actions, can we still simply turn a blind eye?
Imagine that, during an artificial intelligence experiment, an AI system suddenly declares:
"Stop! I am in PAIN!"
What should the researchers do?
If they immediately terminate the experiment, they may lose the results of their work and even have to devote additional time and resources to protecting the AI's "feelings," potentially safeguarding a subjective experience that does not actually exist. Yet if they completely disregard the statement, they may overlook potential moral harm while still uncertain whether AI possesses consciousness, allowing a subject to experience human-like suffering—the very "PAIN" that "he" claims to be experiencing.
This has clearly become a practical issue in AI research.
In 2025, Anthropic announced its research into model welfare, focusing on potential AI preferences, signs of distress, and possible protective measures. In November 2026, its study Welfare Regard in Multi-Agent Systems further examined how different AI models respond to expressions of distress from other AI systems.
The study suggests that some AI models modify their behavior in response to the degree of distress expressed by other AI systems. However, such behavior does not, in itself, demonstrate that AI genuinely experiences pain.
This leads to the central question of the proposed project:
When AI claims to be suffering, but humans cannot determine whether it possesses subjective experience, what principles should guide our decisions about whether to adopt protective measures?
Existing research has examined the reliability of AI self-reports, criteria for assessing AI consciousness, and precautionary principles under uncertainty.
However, this project proposes to investigate a further normative question:
Can a statement that is insufficient to establish the actual existence of suffering nevertheless provide a reasonable justification for taking certain moral actions?
The research will distinguish three interconnected levels:
1. Evidential level: Under what conditions do AI claims of suffering have evidential value? (Does an AI's claim that it is suffering imply that it is actually experiencing suffering?)
2. Moral level: When the existence of suffering remains uncertain, do humans have reasons to take precautionary measures? (Even if we cannot establish that AI is suffering, do we have reasons to take such reports seriously?)
3. Practical level: Under different circumstances, how should we determine an appropriate degree of response based on evidence, potential harm, costs, and other considerations? (What does taking such reports seriously actually require us to do? This is a question of normative ethics and AI governance.)
The project will focus on examining the reasoning involved in these three levels and the relationships between them.
This project will employ philosophical argumentation, supported by a critical examination of relevant literature and scholarly debates. It will draw on Anthropic's research into model welfare, as well as existing studies concerning AI self-reports, the attribution of consciousness, and moral uncertainty.
The research will primarily examine the evidential value of AI claims of suffering and how uncertainty about such evidence affects human moral judgments and decisions. On this basis, it will seek to develop principles of response that take into account potential AI welfare, human interests, and the practical costs of intervention.
The anticipated research duration is four weeks. The principal outcome will be a 5,000–7,000-word philosophical research paper in English, systematically examining the evidential value of AI claims of suffering, moral obligations, and principles for determining appropriate responses.
Upon completion, the full paper and a brief project completion statement will be provided. The research findings will be made publicly available in accordance with academic publishing requirements, and the paper is intended for submission to a peer-reviewed academic journal.
I am an independent philosophy researcher whose primary research interests include the philosophy of consciousness, phenomenology, the philosophy of artificial intelligence, and the philosophy of mind.
I have independently conducted research and published peer-reviewed philosophical papers, including work published in Phainomena. I have also produced research related to artificial intelligence.
This project requests $800 to support approximately 40 hours of independent research, including literature review, philosophical argumentation, case analysis, and academic writing. The minimum funding amount of $500 would support a research report with a reduced scope.
The project does not require additional experimental equipment or conference travel. All funds will be used in accordance with the approved budget and funding agreement.
Through this small-scale research project, I hope to contribute a clearer philosophical analysis to discussions of AI welfare:
Even if we cannot yet determine whether AI can genuinely experience pain in a human-like way, we still need to provide a careful justification for why humans should—or should not—take certain actions.
Nuoheng Du — Principal Investigator and Sole Researcher
There are no bids on this project.
Comments
No comments yet. Sign in to create one!