Summary
An arXiv paper examines how failures occur when AI agents write and execute code, manipulate user interfaces, and transact on their users' behalf. The authors ask what forms these failures take and how they change as agents act with greater autonomy. Drawing on 695 documented incidents from two public repositories, they developed 12 failure modes bottom-up across three interaction stages adapted from Amershi et al.'s guidelines. The excerpt provided is cut off before the full findings on autonomy coding.
AI-assisted summary based on the listed source.
What happened
AI agents write and execute code, manipulate user interfaces, and transact on their users' behalf. We ask what forms failure takes in human-agent interaction, and how those forms change as agents act with greater autonomy. From 695 documented incidents in two public repositories, we developed 12 failure modes...
Why it matters
Teams building or deploying agents that act without step-by-step approval may find the taxonomy useful for sorting where things go wrong. Its focus on how failures change with autonomy suggests that more independent agents need different checks. Readers should review the full paper for the complete autonomy findings, since the excerpt does not include them.
Signal Intelligence
Signal Strength 95%
Technical label SOURCE-BACKED
Public Interest 18
Category RESEARCH
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 0
Novelty Interest Score 48
Consequence Score 18
Curiosity Score 16
Shareability Score 37
