Factors influencing trust in algorithmic decision-making: an indirect scenario-based experiment

Area of Science:

• Decision Science
• Human-Computer Interaction
• Artificial Intelligence Ethics

Background:

• Public distrust in algorithms is prevalent, despite their widespread use in decision-making.
• Educational interventions, such as explaining algorithmic processes, are hypothesized to reduce this distrust.

Purpose of the Study:

• To investigate the relationship between statistical literacy, algorithm explainability, and trust in algorithms.
• To examine how these factors differ across low-stakes and high-stakes decision contexts.

Main Methods:

• A study involving 1,921 participants across 20 countries.
• Analysis of trust in algorithms concerning both low-stakes and high-stakes decisions.
• Assessment of the influence of statistical literacy and algorithm explainability on trust levels.

Main Results:

• Statistical literacy showed a negative association with trust in high-stakes algorithmic decisions.
• Statistical literacy positively correlated with trust in low-stakes scenarios with familiar algorithms.
• Algorithm explainability did not significantly influence participants' trust in algorithms.

Conclusions:

• Statistical literacy empowers individuals to critically assess algorithmic outputs, especially for significant decisions.
• Promoting statistical and artificial intelligence (AI) literacy is crucial for navigating complexities of algorithmic trust.
• Future research should explore user interactions and physiological measures for more accurate trust assessment.