Notes
Contextuality of attitudes towards algorithmic decision-making: Validation and testing of a newly developed universal scale for measuring attitudes towards algorithms (ATAS)
252The development of algorithmic recommendation software and algorithmic decision-making (ADM) has been widely accepted in several fields, from simple recommendation systems in online shops to complex advisory systems in healthcare or criminal justice. While ADM enthusiasts argue that human decision-making could greatly benefit from applying such systems (Lindebaum et al., 2020), research has shown that people do not always appreciate algorithmic advice (Dietvorst et al., 2015). To actually benefit from these powerful tools (e.g. Brzezicki et al., 2020;), it is crucial to understand why and when people do or do not adopt algorithmic advice. Drawing from the Technology Acceptance Model by Davis (1985), we identified attitudes towards algorithms to be a key predictor for the adoption of such systems and effects like algorithm appreciation (Logg et al., 2019) or aversion (Dietvorst et al., 2015). Based on previous research, we hypothesize that attitudes towards algorithms is a dynamic construct influenced by higher-level individual characteristics and contextual factors (Hou & Jung, 2021; Mahmud et al., 2022). As a first step to test this theory, we developed the Attitudes towards Algorithms Scale (ATAS) based on data from two studies (ntotal = 270). In the next step, we validate the scale and test its capability by manipulating contextual factors of an ADM application scenario in a 2x3 between-subject design online study. We used framing and vignettes to describe an application scenario in which a person gets in contact with a hospital that utilizes a comprehensive ADM system. We manipulated criticality of the ADM implementation (how severe are the consequences of an ADM decision) and cruciality for the individual (how much a person is personally affected by an ADM decision) to slightly alter the context. We hypothesize that these alterations affect participants' attitudes towards algorithms and thus result in different ratings on the ATAS subscales.
253Algorithmic decision-making / scale development / contextuality