Kiana Jafari

Gabriela Aránguiz Dias, Kiana Jafari, Allie Griffith et al.

Across healthcare, agentic artificial intelligence (AI) systems are increasingly promoted as capable of autonomous action, yet in practice they currently operate under near-total human oversight due to safety, regulatory, and liability constraints that make autonomous clinical reasoning infeasible in high-stakes environments. While market enthusiasm suggests a revolution in healthcare agents, the conceptual assumptions and accountability structures shaping these systems remain underexamined. We present a qualitative study based on interviews with 20 stakeholders, including developers, implementers, and end users. Our analysis identifies three mutually reinforcing tensions: conceptual fragmentation regarding the definition of `agentic'; an autonomy contradiction where commercial promises exceed operational reality; and an evaluation blind spot that prioritizes technical benchmarks over sociotechnical safety. We argue that agentic {AI} functions as a site of contested meaning-making where technical aspirations, commercial incentives, and clinical constraints intersect, carrying material consequences for patient safety and the distribution of blame.

Kiana Jafari, Paul Ulrich Nikolaus Rust, Duncan Eddy et al.

Learning from human feedback~(LHF) assumes that expert judgments, appropriately aggregated, yield valid ground truth for training and evaluating AI systems. We tested this assumption in mental health, where high safety stakes make expert consensus essential. Three certified psychiatrists independently evaluated LLM-generated responses using a calibrated rubric. Despite similar training and shared instructions, inter-rater reliability was consistently poor ($ICC$ $0.087$--$0.295$), falling below thresholds considered acceptable for consequential assessment. Disagreement was highest on the most safety-critical items. Suicide and self-harm responses produced greater divergence than any other category, and was systematic rather than random. One factor yielded negative reliability (Krippendorff's $α= -0.203$), indicating structured disagreement worse than chance. Qualitative interviews revealed that disagreement reflects coherent but incompatible individual clinical frameworks, safety-first, engagement-centered, and culturally-informed orientations, rather than measurement error. By demonstrating that experts rely on holistic risk heuristics rather than granular factor discrimination, these findings suggest that aggregated labels function as arithmetic compromises that effectively erase grounded professional philosophies. Our results characterize expert disagreement in safety-critical AI as a sociotechnical phenomenon where professional experience introduces sophisticated layers of principled divergence. We discuss implications for reward modeling, safety classification, and evaluation benchmarks, recommending that practitioners shift from consensus-based aggregation to alignment methods that preserve and learn from expert disagreement.

Tanvi Thoria, Kiana Jafari, Marc R. Schlichting et al.

Failures in language model reasoning emerge through distinct processes that leave identifiable signatures in the reasoning trace. We characterize these failures using token-level uncertainty signals, finding they arise through two empirically distinguishable processes. The first is committed failure, in which a model locks onto an incorrect reasoning path early in its trace. A central diagnostic signature is the commitment point, beyond which considering additional tokens hurt rather than help failure detection. In the second, persistent uncertainty, uncertainty instead accumulates throughout, and the full trace is needed to best distinguish failing from successful completions. These signatures reproduce across 23 model-dataset configurations, with the framework's falsifiable predictions holding in 20 of 23 cases, well above chance across both failure modes. Finally, we demonstrate our failure mode framework has direct implications for self-consistency, identifying when uncertainty signals complement it and when it can be selectively skipped. These results offer a foundation for understanding when LLM reasoning failures become detectable and for adapting detection strategies accordingly.