Adaptable Automation Transparency: Should Humans Be Provided Flexibility to Self-Select Transparency Information? ADAPTABLE AUTOMATION TRANSPARENCY

Abstract

Objective We examined whether allowing operators to self-select automation transparency level (adaptable transparency) could improve accuracy of automation use compared to nonadaptable (fixed) low and high transparency. We examined factors underlying higher transparency selection (decision risk, perceived difficulty). Background Increased fixed transparency typically improves automation use accuracy but can increase bias toward agreeing with automated advice. Adaptable transparency may further improve automation use if it increases the perceived expected value of high transparency information. Methods Across two studies, participants completed an uninhabited vehicle (UV) management task where they selected the optimal UV to complete missions. Automation advised the optimal UV but was not always correct. Automation transparency (fixed low, fixed high, adaptable) and decision risk were manipulated within-subjects. Results With adaptable transparency, participants selected higher transparency on 41% of missions and were more likely to select it for missions perceived as more difficult. Decision risk did not impact transparency selection. Increased fixed transparency (low to high) did not benefit automation use accuracy, but reduced decision times. Adaptable transparency did not improve automation use compared to fixed transparency. Conclusion We found no evidence that adaptable transparency improved automation use. Despite a lack of fixed transparency effects in the current study, an aggregated analysis of our work to date using the UV management paradigm indicated that higher fixed transparency improves automation use accuracy, reduces decision time and perceived workload, and increases trust in automation. Application The current study contributes to the emerging evidence-base regarding optimal automation transparency design in the modern workplace.