The Effects of Presenting AI Uncertainty Information on Pharmacists’ Trust in Automated Pill Recognition Technology: Exploratory Mixed Subjects Study
Bibliographic record
Abstract
Background Dispensing errors significantly contribute to adverse drug events, resulting in substantial health care costs and patient harm. Automated pill verification technologies have been developed to aid pharmacists with medication dispensing. However, pharmacists’ trust in such automated technologies remains unexplored. Objective This study aims to investigate pharmacists’ trust in automated pill verification technology designed to support medication dispensing. Methods Thirty licensed pharmacists in the United States performed a web-based simulated pill verification task to determine whether an image of a filled medication bottle matched a known reference image. Participants completed a block of 100 verification trials without any help, and another block of 100 trials with the help of an imperfect artificial intelligence (AI) aid recommending acceptance or rejection of a filled medication bottle. The experiment used a mixed subjects design. The between-subjects factor was the AI aid type, with or without an AI uncertainty plot. The within-subjects factor was the four potential verification outcomes: (1) the AI rejects the incorrect drug, (2) the AI rejects the correct drug, (3) the AI approves the incorrect drug, and (4) the AI approves the correct drug. Participants’ trust in the AI system was measured. Mixed model (generalized linear models) tests were conducted with 2-tailed t tests to compare the means between the 2 AI aid types for each verification outcome. Results Participants had an average trust propensity score of 72 (SD 18.08) out of 100, indicating a positive attitude toward trusting automated technologies. The introduction of an uncertainty plot to the AI aid significantly enhanced pharmacists’ end trust (t28=–1.854; P=.04). Trust dynamics were influenced by AI aid type and verification outcome. Specifically, pharmacists using the AI aid with the uncertainty plot had a significantly larger trust increment when the AI approved the correct drug (t78.98=3.93; P<.001) and a significantly larger trust decrement when the AI approved the incorrect drug (t2939.72=–4.78; P<.001). Intriguingly, the absence of the uncertainty plot led to an increase in trust when the AI correctly rejected an incorrect drug, whereas the presence of the plot resulted in a decrease in trust under the same circumstances (t509.77=–3.96; P<.001). A pronounced “negativity bias” was observed, where the degree of trust reduction when the AI made an error exceeded the trust gain when the AI made a correct decision (z=–11.30; P<.001). Conclusions To the best of our knowledge, this study is the first attempt to examine pharmacists’ trust in automated pill verification technology. Our findings reveal that pharmacists have a favorable disposition toward trusting automation. Moreover, providing uncertainty information about the AI’s recommendation significantly boosts pharmacists’ trust in AI aid, highlighting the importance of developing transparent AI systems within health care.
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 | 0.015 | 0.059 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.004 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.007 | 0.001 |
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".