The DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index.
Cognitive Friction in Clinical Decision Support: A Comparative Study of Judicial and Adjunct Human-AI Interaction Protocols
Open accessAbstract
Artificial intelligence is increasingly used to support clinical decision making, yet concerns remain regarding algorithmic aversion, automation bias and the preservation of meaningful human oversight; while explainable AI aims to improve transparency, less attention has been devoted to the design of human–AI interaction protocols. This study investigates Frictional AI, an interaction paradigm that introduces cognitive friction to encourage critical engagement with AI recommendations. First, semi-structured interviews were conducted with a legal expert and a psychologist and analyzed through thematic analysis to identify legal, ethical, and cognitive requirements for AI-assisted decision support. Second, a user study involving 96 medical residents compared three interaction protocols: a conventional explainable AI-first design (XAI) and two friction-based protocols, namely a judicial protocol based on juxtaposed explanations (Judicial AI, JAI) and an adjunct protocol requiring an initial unsupported decision before AI exposure (AAI). Diagnostic accuracy and confidence, perceived usefulness, completion time, and reliance patterns were evaluated. The interviews highlighted the importance of human-centered explanations, contrastive reasoning, preservation of professional responsibility, and the role of user studies in evaluating human–AI interaction. The quantitative results showed that none of the AI-assisted conditions improved diagnostic accuracy relative to the no-support baseline. However, JAI achieved performance comparable to the baseline, outperforming XAI and AAI, and exhibited the lowest level of over-reliance. Overall findings suggest that the effectiveness of decision-support systems depends not only on model performance and explanation quality but also on interaction design. In conclusion, while preserving diagnostic performance, judicial protocols showed promise in mitigating automation bias and promoting active cognitive engagement in clinical decision support.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
A Unified Approach to Interpreting Model Predictions
Lundberg, S., & Lee, S.-I. (2017) · arXiv · doi:10.48550/arxiv.1705.07874 · Cited by 23,707
Sample Size in Qualitative Interview Studies: Guided by Information Power
Malterud, K., Siersma, V. D., & Guassora, A. D. (2016) · Qualitative Health Research · doi:10.1177/1049732315617444 · Cited by 12,703 · cited 1×
“Although only two experts were interviewed, the sample size is consistent with the concept of information power [46], according to which smaller samples may be sufficient when the study aim is narrow, participants possess highly relevant expertise, and the collected data are rich and focused.”
Confirmation Bias: A Ubiquitous Phenomenon in Many Guises
Nickerson, R. S. (1998) · Review of General Psychology · doi:10.1037/1089-2680.2.2.175 · Cited by 7,267 · cited 3×
“The psychology expert raised concerns that numerical probabilities and complex SHAP plots may be difficult to integrate cognitively and could trigger biases such as confirmation bias [56] or base-rate neglect [33].”
Explanation in artificial intelligence: Insights from the social sciences
Miller, T. (2019) · Artificial Intelligence · doi:10.1016/j.artint.2018.07.007 · Cited by 5,302 · cited 1×
“As discussed by Miller [57], contrastivity is a fundamental characteristic of human reasoning and human-generated explanations and constitutes the foundation of both the established field of counterfactual explanations [58–60] and the more recent field of contrastive explanations [60–62].”
GPOWER: A general power analysis program
Erdfelder, E., Faul, F., & Buchner, A. (1996) · Behavior Research Methods, Instruments, & Computers · doi:10.3758/bf03203630 · Cited by 4,864 · cited 1×
“1) [52] was used to perform sample size calculations.”
Algorithm aversion: People erroneously avoid algorithms after seeing them err.
Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015) · Journal of Experimental Psychology: General · doi:10.1037/xge0000033 · Cited by 3,232 · cited 2×
“This lack of transparency has contributed to widespread skepticism toward AI-assisted decision making, a phenomenon known as algorithmic aversion [6].”
The Honest Broker: Making Sense of Science in Policy and Politics
Pielke, Jr. (2007) · doi:10.1017/cbo9780511818110 · Cited by 2,020
Complacency and Bias in Human Use of Automation: An Attentional Integration
Parasuraman, R., & Manzey, D. H. (2010) · Human Factors: The Journal of the Human Factors and Ergonomics Society · doi:10.1177/0018720810376055 · Cited by 1,710 · cited 4×
“…expert emphasized that responsibility remains primarily with clinicians and healthcare institutions, despite the growing use of AI support and the consequent tendency of automation complacency [23,25] and medical deresponsibilization [26], in which users tend to defer responsibility to machines.”
InfluentialLekadir, K., Frangi, A. F., Porras, A. R., Glocker, B., Cintas, C., Langlotz, C. P., Weicken, E., Asselbergs, F. W., Prior, F., Collins, G. S., et al. (2025) · BMJ · doi:10.1136/bmj-2024-081554 · Cited by 957 · cited 4×
“…governance frameworks that ensure compliance with emerging regulations, robust data governance, continuous risk management, and meaningful human oversight throughout the AI life cycle, such as the European AI Act [9], the NIST AI Risk Management Framework [14], and the FUTURE-AI guidelines [15].”
InfluentialUsing thematic analysis in qualitative research
Ahmed, S. K., Mohammed, R. A., Nashwan, A. J., Ibrahim, R. H., Abdalla, A. Q., M. Ameen, B. M., & Khdhir, R. M. (2025) · Journal of Medicine, Surgery, and Public Health · doi:10.1016/j.glmedi.2025.100198 · Cited by 769 · cited 1×
“Data from the transcriptions of the interviews were analyzed using thematic analysis following the six-phase framework proposed by Braun and Clarke [47].”
What Is the Sense of Agency and Why Does it Matter?
Moore, J. W. (2016) · Frontiers in Psychology · doi:10.3389/fpsyg.2016.01272 · Cited by 645 · cited 2×
“Furthermore, they provide a means to investigate human–AI interaction dynamics that may ultimately lead to undesirable consequences such as automation complacency, deskilling, reduced agency, and derespon-sibilization [23,24,26,28,29].”
A Survey of Methods for Explaining Black Box Models
Guidotti, R., Monreale, A., Ruggieri, S., Turini, F., Giannotti, F., & Pedreschi, D. (2019) · ACM Computing Surveys · doi:10.1145/3236009 · Cited by 629 · cited 1×
“A natural response to these concerns has been the growing adoption of explainable AI methods [7,8].”
Counterfactual explanations and how to find them: literature review and benchmarking
Guidotti, R. (2024) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-022-00831-6 · Cited by 471
Automation bias and verification complexity: a systematic review
Lyell, D., & Coiera, E. (2017) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocw105 · Cited by 451 · cited 3×
“Although algorithmic aversion is commonly regarded as one of the primary barriers to the effective adoption of AI technologies, often framed as a major source of negative technology impact [17], an equally important yet frequently overlooked challenge is automation bias [18].”
Explanations Can Reduce Overreliance on AI Systems During Decision-Making
Vasconcelos, H., Jörke, M., Grunde-McLaughlin, M., Gerstenberg, T., Bernstein, M. S., & Krishna, R. (2023) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3579605 · Cited by 402 · cited 4×
“When pre-support decisions are unavailable, over-reliance and under-reliance have been proposed as practical proxies for automation bias [34,45] and algorithmic aversion, respectively.”
InfluentialStepin, I., Alonso, J. M., Catala, A., & Pereira-Farina, M. (2021) · IEEE Access · doi:10.1109/access.2021.3051315 · Cited by 376
The Alignment Problem: Machine Learning and Human Values
Christian, B. (2021) · Perspectives on Science and Christian Faith · doi:10.56315/pscf12-21christian · Cited by 311
Chou, Y.-L., Moreira, C., Bruza, P., Ouyang, C., & Jorge, J. (2022) · Information Fusion · doi:10.1016/j.inffus.2021.11.003 · Cited by 302 · cited 1×
“As discussed by Miller [57], contrastivity is a fundamental characteristic of human reasoning and human-generated explanations and constitutes the foundation of both the established field of counterfactual explanations [58–60] and the more recent field of contrastive explanations [60–62].”
Artificial intelligence risk management framework :
National Institute of Standards and Technology (US) (2024) · NIST Trustworthy and Responsible AI · doi:10.6028/nist.ai.600-1 · Cited by 274
Explainable AI: current status and future directionsPreprint
Gohel, P., Singh, P., & Mohanty, M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.07045 · Cited by 248
Aquino, Y. S. J., Rogers, W. A., Braunack-Mayer, A., Frazer, H., Win, K. T., Houssami, N., Degeling, C., Semsarian, C., & Carter, S. M. (2023) · International Journal of Medical Informatics · doi:10.1016/j.ijmedinf.2022.104903 · Cited by 131 · cited 2×
“Finally, excessive dependence on AI may promote deskilling [28], limiting both the maintenance and, in some cases, the development of professional expertise [29].”
How AI Impacts Skill Formation
Shen, J. H., & Tamkin, A. (2026) · arXiv · doi:10.48550/arxiv.2601.20245 · Cited by 120
Rams, hounds and white boxes: investigating human–AI collaboration protocols in medical diagnosis
Cabitza, F., Campagner, A., Ronzio, L., Cameli, M., Mandoli, G. E., Pastore, M. C., Sconfienza, L. M., Folgado, D., Barandas, M., & Gamboa, H. (2023) · Artificial Intelligence in Medicine · doi:10.1016/j.artmed.2023.102506 · Cited by 117 · cited 6×
“Nevertheless, these findings are consistent with another previous observation by Cabitza et al. [41] as discussed in Section 1.2 and align with similar evidence reported in the literature.”
InfluentialCao, S., Liu, A., & Huang, C. (2024) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3637318 · Cited by 88 · cited 1×
“Examples in healthcare include the works of Cao et al. [35], Bashkirova et al. [36], and Buçinca et al. [37].”
The accuracy of effect-size estimates under normals and contaminated normals in meta-analysis
Marfo, P., & Okyere, G. A. (2019) · Heliyon · doi:10.1016/j.heliyon.2019.e01838 · Cited by 84 · cited 1×
“Effect sizes comparing the intervention conditions (XAI, JAI, and AAI) with the control condition (NoAI) were computed following established practice [38,53].”
Nicora, G., Pe, S., Santangelo, G., Billeci, L., Aprile, I. G., Germanotta, M., Bellazzi, R., Parimbelli, E., & Quaglini, S. (2025) · Journal of NeuroEngineering and Rehabilitation · doi:10.1186/s12984-025-01605-z · Cited by 77
Self-citationBuçinca, Z., Swaroop, S., Paluch, A. E., Doshi‐Velez, F., & Gajos, K. Z. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713229 · Cited by 75 · cited 4×
“For example, Buçinca et al. [37] compared multiple interaction protocols, including an adjunct design in which users first made an independent decision and subsequently received a contrastive explanation [60–62] comparing the model prediction with the first user’s decision (as the foil).”
InfluentialNever tell me the odds: Investigating pro-hoc explanations in medical decision making
Cabitza, F., Natali, C., Famiglini, L., Campagner, A., Caccavella, V. M., & Gallazzi, E. (2024) · Artificial Intelligence in Medicine · doi:10.1016/j.artmed.2024.102819 · Cited by 58 · cited 10×
“In agreement with both previous studies [22,38] and the findings reported in Section 3.2.2, additional work is also needed to investigate how cognitive friction can be personalized according to user characteristics and contextual factors.”
InfluentialBashkirova, A., & Krpan, D. (2024) · Computers in Human Behavior: Artificial Humans · doi:10.1016/j.chbah.2024.100066 · Cited by 55 · cited 1×
“Examples in healthcare include the works of Cao et al. [35], Bashkirova et al. [36], and Buçinca et al. [37].”
Cabitza, F., Campagner, A., Angius, R., Natali, C., & Reverberi, C. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581095 · Cited by 51 · cited 7×
“Automation bias is typically defined as a detrimental over-reliance on DSSs, whereby users defer their judgment to the machine even when independent verification would be warranted [17,18].”
InfluentialContrastive Explanations In Neural Networks
Prabhushankar, M., Kwon, G., Temel, D., & AlRegib, G. (2020) · Proceedings - International Conference on Image Processing · doi:10.1109/icip40778.2020.9190927 · Cited by 47
Grote, T., & Berens, P. (2021) · Bioethics · doi:10.1111/bioe.12957 · Cited by 44 · cited 1×
“In this respect, the phenomenon may also be related to defensive medicine [27], whereby clinicians may defer to AI recommendations or modify their clinical decisions primarily to reduce legal liability or professional criticism.”
Comparing Alternative Corrections for Bias in the Bias-Corrected Bootstrap Test of Mediation
Chen, D., & Fritz, M. S. (2021) · Evaluation & the Health Professions · doi:10.1177/01632787211024356 · Cited by 44 · cited 1×
“95%- confidence intervals were estimated using bias-corrected and accelerated (BCa) bootstrap resampling [54].”
Bergomi, L., Nicora, G., Orlowska, M. A., Podrecca, C., Bellazzi, R., Fregosi, C., Salinaro, F., Bonzano, M., Crescenzi, G., Speciale, F., et al. (2025) · BMC Medical Informatics and Decision Making · doi:10.1186/s12911-025-03045-0 · Cited by 33 · cited 3×
“In preceding studies [43,49], radiomic features were extracted from the images and associated with personal and clinical data obtained from each patient’s electronic health record.”
Self-citationThe key role of differential diagnosis in diagnosis
Jain, B. P. (2017) · Diagnosis · doi:10.1515/dx-2017-0005 · Cited by 26 · cited 1×
“This design resembles the differential diagnosis framework [42] widely used in healthcare, and may therefore represent a promising strategy for mitigating automation bias while preserving users’ sense of agency, responsibility, and critical judgment.”
Cognitive biases in internal medicine: a scoping review
Loncharich, M. F., Robbins, R. C., Durning, S. J., Soh, M. Y., & Merkebu, J. (2023) · Diagnosis · doi:10.1515/dx-2022-0120 · Cited by 24 · cited 2×
“This phenomenon is well documented in healthcare [64] and was explicitly highlighted both in the literature [33] and during the expert interviews (Section 3.1) as a potential risk associated with AI-assisted decision support.”
InfluentialExplanatory Paradigms in Neural Networks: Towards relevant and contextual explanations
AlRegib, G., & Prabhushankar, M. (2022) · IEEE Signal Processing Magazine · doi:10.1109/msp.2022.3163871 · Cited by 22 · cited 1×
“In this respect, AlRegib and Prabhushankar [63] argue that classical, contrastive, and counterfactual explanations can contribute together to the generation of “abductive” evidence, a concept closely related to the differential diagnosis process routinely employed by clinicians in everyday practice.”
Dissenting Explanations: Leveraging Disagreement to Reduce Model Overreliance
Reingold, O., Shen, J. H., & Talati, A. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i19.30151 · Cited by 18
Lesley, U., & Kuratomi, A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-66535-6_32 · Cited by 18 · cited 1×
“Is the “Clinical Explanation Satisfaction Scale” [66] clear and easy to understand?”
Cabitza, F., Famiglini, L., Fregosi, C., Pe, S., Parimbelli, E., La Maida, G. A., & Gallazzi, E. (2025) · Proceedings of the 30th International Conference on Intelligent User Interfaces · doi:10.1145/3708359.3712157 · Cited by 11 · cited 10×
“…considerations discussed in Section 1.2; while previous studies have shown that adjunct-based interaction protocols may negatively affect decision quality [41], Cabitza et al. [38] demonstrated that combining adjunct protocols with judicial principles can produce more favorable outcomes.”
InfluentialSelf-citationCatalano, M., Bortolotto, C., Nicora, G., Achilli, M. F., Consonni, A., Ruongo, L., Callea, G., Tito, A. L., Biasibetti, C., Donatelli, A., et al. (2023) · European Journal of Radiology Open · doi:10.1016/j.ejro.2023.100497 · Cited by 10 · cited 2×
“The study was conducted within the context of the ALFABETO project [43,44].”
Self-citationDin, S. U., Kemna, R., Ket, J. C. F., Iqbal, M., Bohoudi, O., Hoogendoorn, M., Beretta, E., & Lisowska, A. (2026) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2026.1819422 · Cited by 8 · cited 1×
“Despite receiving relatively limited attention in the literature [21], automation bias has been widely documented in healthcare settings [22,23].”
Pe, S., Famiglini, L., Gallazzi, E., Bortolotto, C., Carone, L., Cisarri, A., Salina, A., Preda, L., Bellazzi, R., Cabitza, F., et al. (2024) · Methods of Information in Medicine · doi:10.1055/a-2562-2163 · Cited by 7 · cited 1×
“Specifically, the Grad-CAM algorithm [40] was employed as an example for juxtaposed evidence generation: class activation maps highlight image regions that contributed most strongly to a given prediction; by producing separate maps for both the fracture and non-fracture classes, the system was able…”
Self-citationRight, No Matter Why: AI Fact-checking and AI Authority in Health-related Inquiry SettingsPreprint
Sergeeva, E., Sergeeva, A., Tang, H., Bongard-Blanchy, K., & Szolovits, P. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2310.14358 · Cited by 5
Sureau, C. (1995) · Journal of Assisted Reproduction and Genetics · doi:10.1007/bf02212919 · Cited by 4 · cited 4×
“…expert emphasized that responsibility remains primarily with clinicians and healthcare institutions, despite the growing use of AI support and the consequent tendency of automation complacency [23,25] and medical deresponsibilization [26], in which users tend to defer responsibility to machines.”
InfluentialA Comprehensive Analysis and Prediction of Diabetes Diseases Using Deep Learning Techniques
Tripathy, N., Nayak, S. K., Addula, S. R., et al. (2025) · CRC Press eBooks · doi:10.1201/9781003531050-4 · Cited by 1
Research works printed but not linked: 5
- IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems (2019). Ethically Aligned Design: A Vision for Prioritizing Human Well-Being with Autonomous and Intelligent Systems, IEEE Standards Association.
- Natali, C. (2023). Per Aspera ad Astra, or Flourishing via Friction: Stimulating Cognitive Activation by Design through Frictional Decision Support Systems. CHItaly-DC 2023 CHItaly 2023 Doctoral Consortium CHItaly 2023 Proceedings of the Doctoral Consortium of the 15th Biannual Conference of the Italian SIGCHI Chapter (CHItaly 2023), CEUR-WS.
- Norman, D.A. (2011). Living with Complexity, MIT Press.
- Nicora, G., Tito, A.L., Donatelli, A., Callea, G., Biasibetti, C., Galli, M.V., Comotto, F., Bortolotto, C., Perlini, S., and Preda, L. (2021, January 29). ALFABETO: Supporting COVID-19 hospital admissions with Bayesian Networks. Proceedings of the Workshop on Towards Smarter Health Care: Can Artificial Intelligence Help? co-located with 20th International Conference of the Italian Association for Artificial Intelligence (AIxIA2021), Online Event.
- Moens, M.F., Huang, X., Specia, L., and Yih, S.W.T. (2021). Contrastive Explanations for Model Interpretability. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics.
Metadata
- Authors
- Pe, S., Bergomi, L., Nicora, G., Simonelli, C. A., Kour, P., Diaz, E., Alendal, G., Hernáiz Ferrer, A. I., Corso, V., Bortolotto, C., Zuccaro, V., Salinaro, F., Preda, L., & Parimbelli, E.
- Affiliations
- University of Pavia (IT)
- University of Bergen (NO)
- Department of Global Public Health and Primary Care, Faculty of Medicine, University of Bergen, 5009 Bergen, Norway
- Department of Radiology, IRCCS Policlinic San Matteo Foundation, 27100 Pavia, Italy
- University of Ottawa (CA)
- Corresponding author
- Pe, Samuele
- Parimbelli, Enea
- ORCID
- Venue
- Machine Learning and Knowledge Extraction
- Publisher
- MDPI AG
- Published
- 2026-07-22
- Volume
- 8
- Issue
- 7
- Pages
- 216
- ISSN
- 2504-4990 (ISSN-L)
- OpenAlex
- W7170090355
- Cited by
- 0
- References
- 42
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Machine Learning and Knowledge Extraction (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- Funders
- Italian Ministry of Research (PNC0000007)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Social Sciences › Human Factors and Ergonomics › Human-Automation Interaction and Safety
- Keywords
- clinical decision support, human-AI interaction, cognitive friction, explainable AI, automation bias, algorithmic aversion, overreliance on AI, diagnostic accuracy, medical residents, interaction design, contrastive reasoning, professional responsibility
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.