ECP 2022 A

Trust in Algorithms and Statistical Literacy in Low- and High-Stake Contexts

Asil Ali Özdoğru and Merve Büşra Çetin

Many organizations and governments around the world are collecting large amounts of data and using various algorithms to process data and make decisions. These algorithms have important consequences as they are trusted to assist decision making in many contexts. Individuals’ trust in algorithms can be shaped by many personal and contextual factors. Since algorithms have a statistical nature and different levels of explainability, statistical literacy, which is individuals’ ability to understand statistical and probabilistic information, may play a role in their trust in algorithms. Individuals’ trust can also show variation depending on the context of decisions. Low-stake situations such as a restaurant reservation and high-stakes situations such as employment decisions can have different effects on trust. This study tested the effects of statistical literacy, degree of explainability, and decision context on Turkish university students’ level of trust in algorithms. A total of 110 students (95 females, 14 males, 1 unknown) from a foundation university in İstanbul participated in the study. Participants responded to an online survey including basic demographic questions, six items on trust in algorithms, 14 items on statistical literacy, and two items about 12 scenarios featuring algorithms with different levels of explainability and context. Half of the participants (n = 52) received low-stake scenarios with explainability and high-stake scenarios without explainability, while others (n = 58) received low-stake scenarios without explainability and high-stake scenarios with explainability. There was no statistically significant correlation between trust in algorithms and statistical literacy scores. Trust in algorithm ratings in the scenarios did not significantly differ between explainability conditions. On the other hand, trust ratings in high-stake scenarios were significantly lower than the low-stake ones. Results show that individuals’ tendency to trust in algorithms is not related to their levels of statistical literacy but related to context of algorithms in use.

Citation: Özdoğru, A. A., & Çetin, M. B. (2022, July 5-8). Trust in algorithms and statistical literacy in low- and high-stake contexts [Conference presentation]. 17th European Congress of Psychology, Ljubljana, Slovenia. https://www.ecp2022.eu

Media: Üsküdar Üniversitesi Psikoloji Bölümü Çalışmalarıyla Avrupa Psikoloji Kongresinde! (2022, Temmuz 7). Üsküdar Haber Ajansı. https://uskudar.edu.tr/tr/icerik/8143/uskudar-universitesi-psikoloji-bolumu-calismalariyla-avrupa-psikoloji-kongresinde