Predictions in a Noisy Environment: The Influence of Automation Reliability and Transparency on User Trust

This study examined the potential interaction between automation reliability and automation transparency in the context of a weather-related decision to close a school in the face of a forecast blizzard. Participants received snowfall and closure advice from four different weather services dependent upon an automation weather model that differed in its degree of reliability (100% vs. 67%). Two of the services simply provided advice, and for two of them, this advice was accompanied by transparency in the form of the raw data, or weather symptoms display that was input into the weather model. The results revealed that degraded reliability of the automation lowered trust in and dependence on the advice and hurt overall accuracy. In a significant interaction, these degraded effects were mitigated by automation transparency. These findings were interpreted in terms of a mechanism of optimal attention allocated (weighting) between the raw data transparency and the weather service advice.