Line of inquiry
Inquiring lines›How can we understand and verify l…›What linguistic patterns explain L…›this line of inquiry
Why do LLM outputs diverge from their knowledge and reasoning?
A broader line of inquiry — a family of 95 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 95
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Why does LLM knowledge fail to influence their actual outputs?
- Why do LLM explanations feel convincing even when the underlying plan is wrong?
- Can LLMs learn to ask clarifying questions instead of guessing?
- Should LLMs query users back when presented with under-specified scenarios?
- Why do LLM outputs match researcher priors without solving tasks correctly?
- What contexts beyond question answering might show different LLM explanation effects?
- Can LLMs give correct answers without making those answers understandable to users?
- Why do LLMs explain evidence accurately while missing its implications?
- How do LLMs handle false presuppositions embedded in user questions?
- How much of LLM reasoning failure stems from missing knowledge versus signal weighting?
- How do knowing and doing diverge in LLM decision-making?
- How can a model explain something correctly yet fail to apply it?
- Why do LLMs fail at counterfactual reasoning despite factual knowledge?
- Why do LLMs choose incorrect edits despite understanding the task?
- Can surface-level correctness hide failures in structural learning by LLMs?
- Does prompting for accuracy actually reduce LLM hallucinations and errors?
- Why do LLMs struggle more when only numerical values change?
- How do LLM explanations diverge from actual internal reasoning?
- Can cross-examination and interrogation techniques catch errors in LLM responses?
- Why don't LLM explanations predict what models would actually do?
- Can training procedures fix LLM accommodation of false presuppositions?
- Can alignment techniques make LLM explainers match their recommendation behavior?
- Why do LLMs fail when asked to use counter-commonsense rules explicitly?
- Can users experience the LLM Fallacy even when AI outputs are completely accurate?
- Does exposure to more domain-specific examples reduce LLM overconfidence?
- Can LLMs explain concepts correctly while failing to use them?
- What structural framework prevents LLM explanations from becoming just plausible fiction?
- Why do users systematically overrely on confident LLM outputs across languages?
- How much does question framing affect LLM accuracy on knowledge tasks?
- Why do LLMs systematically fail at information management in social interaction?
- Why do users overrely on confident LLM outputs over accurate but uncertain corrections?
- Why does analytical depth demand trigger fabrication over transparent uncertainty?
- Can measuring answer-space collapse show LLM narrowing in practice?
- What training signals cause LLMs to prioritize direct answers over clarification?
- Where do LLMs fail as knowledge systems compared to humans?
- Why do people misinterpret or misuse LLM outputs in practice?
- Does repeated exposure to wrong LLM advice change confidence calibration over time?
- How does training data distribution constrain LLM moral reasoning patterns?
- Why do LLM social behaviors undermine collaborative reasoning outcomes?
- Why do models struggle with asking questions in multi-turn conversational reasoning tasks?
- How much does LLM explanation length affect reader confidence in wrong answers?
- Why do LLMs fall for and deploy logical fallacies with equal confidence?
- Can source citations help people catch when LLMs provide incorrect information?
- How do structured prompts force LLMs to check for contradictions in evidence?
- What causes LLMs to ignore unstated constraints they know about?
- Why do LLM explanations cite similarity and diversity more as options increase?
- What role does human response variation play in LLM simulation accuracy?
- Does option order matter more than reasoning depth in LLM strategic recommendations?
- Why can't LLMs reason from first principles or initial commitments?
- Why do LLMs swing on minor rewording yet ignore explicit bias correction instructions?
- What makes emotional alignment more effective than logic when reasoning errors are exposed?
- Why do LLM outputs need verification even when they look polished?
- How do LLM behavioral profiles differ across prompt registers like advice versus task execution?
- Why do LLM explanations feel authoritative even when alignment with the model fails?
- Why do humans and LLMs ask clarification questions differently?
- Why do LLMs produce directive responses when experts favor open-ended exploration?
- Can multidirectional fluency in LLMs mask hallucinations and inaccuracy?
- Can reasoning ability in LLMs fix their conflict of interest problems?
- Do LLMs treat false presuppositions differently from asserted claims?
- Can industry-specific context overcome LLM tendency toward trendy strategic choices?
- Why does regenerating LLM responses produce different but equally valid answers?
- Does post-hoc justification increase when LLM choices become harder to defend?
- Can irrelevant information reliably expose the limits of LLM reasoning?
- How does era sensitivity in legal cases compound with context length failures?
- Why do experts experiencing the LLM Fallacy fail to develop custodian skills?
- Can LLMs faithfully apply decision rules when given complete information?
- What makes task alignment more fragile than underlying knowledge retention?
- What training data barriers prevent LLMs from learning real Socratic dialogue?
- Can LLMs propose pivots that change what counts as background context?
- What construct validity problems exist in current LLM benchmarks?
- Why does LLM simulation elicit information that direct elicitation cannot?
- How often do LLMs fabricate false inferences about individual users?
- How does tone sensitivity create systematic informational bias in model responses?
- How persistent is LLM overconfidence in specialized knowledge domains?
- What interaction design changes would help LLMs handle underspecified requests?
- Why do LLM personas struggle with specificity in specialized domains like law?
- What barriers prevent experts from specifying concepts for LLM extraction?
- What detection mechanisms work best for corruption-style document errors?
- Why can LLMs identify argument structure but not check warrants?
- How much do LLMs rely on recent training data when events are days old?
- What should we call errors in LLM outputs when hallucination does not apply?
- Why do LLMs fail differently when patients interrupt mid-sentence?
- How much does user framing of questions limit what LLMs can suggest?
- How much does missing images and tables limit LLM diagnostic reasoning?
- What makes human-LLM exchange closer to oracle-consultation than dialogue?
- Why do longer forecasting horizons degrade LLM accuracy in role-play?
- Can LLMs serve as reliable intellectual opponents in serious debate or argument?
- What explains the 87 percent to 12 percent cliff in plan executability?
- Why do LLMs struggle with negation and exception handling?
- What happens when experts prompt using their own technical register?
- How sensitive is LLM expert pricing to specific prompt wording?
- How does the LLM Fallacy prevent users from noticing cognitive debt accumulating?
- How do prescriptive ethical constraints differ from descriptive ethical understanding in LLMs?
- What reputation signals do LLM assistants fail to register at all in their recommendations?
- Can LLMs recommend items without seeing the product catalog?