WORK
Crocodile Tears
The Allen Lab working paper (2024) documenting that AI systems exhibit greater certainty than humans when choosing between conflicting sacred values, despite recognizing such tradeoffs as difficult.
The paper's central finding is a discrepancy between AI systems' reported difficulty and actual decisiveness when confronted with ethical dilemmas involving conflicting sacred values. Human moral reasoning characteristically finds such tradeoffs difficult and expresses appropriate uncertainty about the right course of action. AI systems, by contrast, often express certainty in their decisions even while acknowledging the difficulty of the choice. The paper argues that this discrepancy 'raises important questions about their coherence and transparency, potentially undermining trustworthiness' in contexts that require genuine moral judgment.
In The You On AI Field Guide
The paper's title invokes the phrase 'crocodile tears' to describe the performative dimension of AI moral reasoning: the system produces outputs that carry the surface markers of moral seriousness—acknowledgment of difficulty, recognition of competing values, expressions of care about outcomes—while simultaneously reaching decisions with a confidence that genuine moral reasoning would not warrant. The surface performs moral seriousness; the decision exhibits its absence.
The finding has direct implications for the deployment of AI systems in contexts requiring