In AI We Trust: Ethics, Artificial Intelligence, and Reliability

Area of Science:

• Philosophy of Technology
• Artificial Intelligence Ethics
• Human-Computer Interaction

Background:

• Assessing artificial intelligence (AI) is complicated by human tendency to anthropomorphize.
• Attributing human moral activities, such as trust, to AI is particularly problematic.
• The European Commission's High-level Expert Group on AI (HLEG) advocates for trustworthy AI.

Purpose of the Study:

• To critically examine the claim that artificial intelligence (AI) can be trusted.
• To analyze AI's capacity for trust based on prevalent philosophical definitions.
• To differentiate between genuine trust and reliance in the context of AI.

Main Methods:

• Analysis of affective, normative, and rational accounts of trust.
• Evaluation of AI's capabilities against the requirements of each trust account.
• Philosophical argumentation regarding the nature of trust and responsibility.

Main Results:

• AI lacks the emotive states and accountability required for affective and normative trust.
• AI meets the criteria for the rational account of trust, but this constitutes reliance, not trust.
• Viewing AI as trustworthy undermines human interpersonal trust and accountability.

Conclusions:

• Artificial intelligence (AI) cannot be genuinely trusted based on established definitions.
• The concept of 'trustworthy AI' is a misapplication of the term, conflating trust with reliance.
• It is crucial to avoid anthropomorphizing AI and to maintain clear lines of responsibility for AI development and deployment.