Appropriate trust in artificial intelligence for the optical diagnosis of colorectal polyps: The role of human/artificial intelligence interaction.

BACKGROUND AND AIMS: Computer-aided diagnosis (CADx) for the optical diagnosis of colorectal polyps is thoroughly investigated. However, studies on human-artificial intelligence interaction are lacking. Our aim was to investigate endoscopists' trust in CADx by evaluating whether communicating a calibrated algorithm confidence score improved trust. METHODS: Endoscopists optically diagnosed 60 colorectal polyps. Initially, endoscopists diagnosed the polyps without CADx assistance (initial diagnosis). Immediately afterward, the same polyp was again shown with a CADx prediction: either only a prediction (benign or premalignant) or a prediction accompanied by a calibrated confidence score (0-100). A confidence score of 0 indicated a benign prediction, 100 a (pre)malignant prediction. In half of the polyps, CADx was mandatory, and for the other half, CADx was optional. After reviewing the CADx prediction, endoscopists made a final diagnosis. Histopathology was used as the gold standard. Endoscopists' trust in CADx was measured as CADx prediction utilization: the willingness to follow CADx predictions when the endoscopists initially disagreed with the CADx prediction. RESULTS: Twenty-three endoscopists participated. Presenting CADx predictions increased the endoscopists' diagnostic accuracy (69.3% initial vs 76.6% final diagnosis, P < .001). The CADx prediction was used in 36.5% (n = 183 of 501) disagreements. Adding a confidence score led to lower CADx prediction utilization, except when the confidence score surpassed 60. Mandatory CADx decreased CADx prediction utilization compared to optional CADx. Appropriate trust-using correct or disregarding incorrect CADx predictions-was 48.7% (n = 244 of 501). CONCLUSIONS: Appropriate trust was common, and CADx prediction utilization was highest for the optional CADx without confidence scores. These results express the importance of a better understanding of human-artificial intelligence interaction.

The Impact of Expectation Management and Model Transparency on Radiologists' Trust and Utilization of AI Recommendations for Lung Nodule Assessment on Computed Tomography: Simulated Use Study.

BACKGROUND: Many promising artificial intelligence (AI) and computer-aided detection and diagnosis systems have been developed, but few have been successfully integrated into clinical practice. This is partially owing to a lack of user-centered design of AI-based computer-aided detection or diagnosis (AI-CAD) systems. OBJECTIVE: We aimed to assess the impact of different onboarding tutorials and levels of AI model explainability on radiologists' trust in AI and the use of AI recommendations in lung nodule assessment on computed tomography (CT) scans. METHODS: In total, 20 radiologists from 7 Dutch medical centers performed lung nodule assessment on CT scans under different conditions in a simulated use study as part of a 2×2 repeated-measures quasi-experimental design. Two types of AI onboarding tutorials (reflective vs informative) and 2 levels of AI output (black box vs explainable) were designed. The radiologists first received an onboarding tutorial that was either informative or reflective. Subsequently, each radiologist assessed 7 CT scans, first without AI recommendations. AI recommendations were shown to the radiologist, and they could adjust their initial assessment. Half of the participants received the recommendations via black box AI output and half received explainable AI output. Mental model and psychological trust were measured before onboarding, after onboarding, and after assessing the 7 CT scans. We recorded whether radiologists changed their assessment on found nodules, malignancy prediction, and follow-up advice for each CT assessment. In addition, we analyzed whether radiologists' trust in their assessments had changed based on the AI recommendations. RESULTS: Both variations of onboarding tutorials resulted in a significantly improved mental model of the AI-CAD system (informative P=.01 and reflective P=.01). After using AI-CAD, psychological trust significantly decreased for the group with explainable AI output (P=.02). On the basis of the AI recommendations, radiologists changed the number of reported nodules in 27 of 140 assessments, malignancy prediction in 32 of 140 assessments, and follow-up advice in 12 of 140 assessments. The changes were mostly an increased number of reported nodules, a higher estimated probability of malignancy, and earlier follow-up. The radiologists' confidence in their found nodules changed in 82 of 140 assessments, in their estimated probability of malignancy in 50 of 140 assessments, and in their follow-up advice in 28 of 140 assessments. These changes were predominantly increases in confidence. The number of changed assessments and radiologists' confidence did not significantly differ between the groups that received different onboarding tutorials and AI outputs. CONCLUSIONS: Onboarding tutorials help radiologists gain a better understanding of AI-CAD and facilitate the formation of a correct mental model. If AI explanations do not consistently substantiate the probability of malignancy across patient cases, radiologists' trust in the AI-CAD system can be impaired. Radiologists' confidence in their assessments was improved by using the AI recommendations.

Determinants of Domestic Risk Prevention Behavior: The Importance of Separating Effects Within-Persons and Between-Persons.

The effects of vulnerability, severity, costs, effort, and effectiveness on prevention behavior, derived from protection motivation theory and the health belief model, have been extensively tested in the literature and have all been shown to predict rather well. In this study we test the effects of these determinants in a new context: the domestic risk prevention domain. The specific behaviors under study are related to the risks of burglary, fire, and water damage. In addition to previous studies, our multilevel research design allows us to evaluate which differences in the performance of domestic prevention behavior can be attributed to differences between persons and which to differences between behaviors within persons. Our results show that all determinants are relevant predictors for domestic risk prevention behavior. Disentangling the within-person and between-person effects shows that prevention behavior depends more on the relative evaluation of the prevention behavior determinants for a given person (e.g., a person perceives a smoke alarm to be more effective than antiburglar strips), than on the differences between persons regarding the general perception of these determinants (e.g., some persons find prevention behaviors in general more effective than other persons). To increase the performance of domestic risk prevention behaviors, we advise that interventions should focus on increasing a person's perception of risks and prevention behaviors relative to other risks and prevention behaviors rather than focusing on changing people's general perceptions of all risks and behaviors or focusing on specific target groups.

Playing with fire. Understanding how experiencing a fire in an immersive virtual environment affects prevention behavior.

A potentially effective way to influence people's fire prevention behavior is letting them experience a fire in an immersive virtual environment (IVE). We analyze the effects of experiencing a fire in an IVE (versus an information sheet) on psychological determinants of behavior-knowledge, vulnerability, severity, self-efficacy, and locus of control-based mainly on arguments from Protection Motivation Theory and the Health Belief Model. Crucial in our setup is that we also relate these determinants to actual prevention behavior. Results show that IVE has the hypothesized effects on vulnerability, severity, and self-efficacy, and an unexpected negative effect on knowledge. Only knowledge and vulnerability showed subsequent indirect effects on actual prevention behavior. There remains a direct positive effect of IVE on prevention behavior that cannot be explained by any of the determinants. Our results contradict the implicit assumption that an induced change in these psychological determinants by IVE, necessarily implies a change in behavior. A recommendation for research on the effects of IVE's is, whenever possible, to study the actual target behavior as well.

Development of a non-melanoma skin cancer detection model.

BACKGROUND: The incidence and prevalence of skin cancer is rising. A detection model could support the (screening) process of diagnosing non-melanoma skin cancer. METHODS: A questionnaire was developed containing potential actinic keratosis (AK) and basal cell carcinoma (BCC) characteristics. Three nurses diagnosed 204 patients with a lesion suspicious of skin (pre)malignancy and filled in the questionnaire. Logistic regression analyses generated prediction models for AK and BCC. RESULTS: A prediction model containing nine characteristics correctly predicted the presence or absence of AK in 83.2% of the cases. BCC was predicted correctly in 91.4% of the cases by a model containing eight characteristics. The nurses correctly diagnosed AK in 88.3% and BCC in 90.9% of the cases. CONCLUSIONS: Detection or screening models for AK and BCC could be made with a limited number of variables. Nurses also diagnosed skin lesions correctly in a high percentage of cases. Further research is necessary to investigate the robustness of these findings, whether the percentage of correct diagnoses can be improved and how best to implement model-based prediction in the diagnostic process.