Impacts of cognitive forcing and need for cognition on biased AI-assisted decision making about mental health emergencies
Bibliographic record
Abstract
Artificial intelligence (AI) trained to predict psychiatric inpatient violence may overestimate risks for marginalized groups, making it critical to find ways to mitigate reliance on biased AI in this context. One potential solution is Cognitive forcing (CF), or interventions that delay AI information or slow the decision-making process. Benefits of CF may be modulated by traits, such as Need for Cognition (NFC), or the tendency to engage with complex, cognitive tasks. To examine how CF and NFC impact AI-assisted decision-making about violence risk, we conducted two experiments. In Experiment 1, participants (n = 281) made decisions about violence risk based on vignettes describing various patients experiencing mental health emergencies, and they were randomized to view biased or unbiased AI recommendations. In Experiment 2, participants (n = 373) made similar decisions, and they were randomized to view biased AI recommendations with one of three CF interventions or no CF. All participants completed measures of NFC. In both experiments, participants made biased decisions (overestimating violence risk for marginalized patients) when viewing biased AI recommendations. In Experiment 2, CF interventions did not mitigate this decision-making bias; however, participants reporting high NFC were less likely to make biased decisions when viewing biased AI recommendations, compared to those with low NFC. CF may not effectively safeguard against the impact of biased AI in high-stakes settings, like acute mental health care, or for decisions about violence risk prediction, which are fraught with social or racial stereotypes. However, trait NFC may mitigate reliance on biased AI information, highlighting a role of psychological factors. Further research is needed into various factors that promote equitable AI-assisted decision-making for mental health.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.053 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".