2021/03/02 by Alexandra Zytek, Dongyu Liu, Zytek, Alexandra +5 · 4 citations
Computer Science · Psychology · #Data Visualization and Analytics #FOS: Computer and information sciences #Human-Computer Interaction (cs.HC) #Machine Learning (cs.LG) #Mental Health Research Topics
paper · pdf · doi:10.48550/arxiv.2103.02071
openalex publication_date 2021/03/02 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Machine learning (ML) is being applied to a diverse and ever-growing set of\ndomains. In many cases, domain experts - who often have no expertise in ML or\ndata science - are asked to use ML predictions to make high-stakes decisions.\nMultiple ML usability challenges can appear as result, such as lack of user\ntrust in the model, inability to reconcile human-ML disagreement, and ethical\nconcerns about oversimplification of complex problems to a single algorithm\noutput. In this paper, we investigate the ML usability challenges that present\nin the domain of child welfare screening through a series of collaborations\nwith child welfare screeners. Following the iterative design process between\nthe ML scientists, visualization researchers, and domain experts (child\nscreeners), we first identified four key ML challenges and honed in on one\npromising explainable ML technique to address them (local factor\ncontributions). Then we implemented and evaluated our visual analytics tool,\nSibyl, to increase the interpretability and interactivity of local factor\ncontributions. The effectiveness of our tool is demonstrated by two formal user\nstudies with 12 non-expert participants and 13 expert participants\nrespectively. Valuable feedback was collected, from which we composed a list of\ndesign implications as a useful guideline for researchers who aim to develop an\ninterpretable and interactive visualization tool for ML prediction models\ndeployed for child welfare screeners and other similar domain experts.\n