The role of algorithmic accountability in shaping developers’ deontological judgements and implementation behaviours

Organisations increasingly delegate consequential judgements to algorithmic decision-making systems (ADMS)—software that renders decisions with minimal human intervention, as in credit scoring and fraud detection. Such systems can produce unethical outcomes that cause substantial reputational and societal harm, including discrimination against minorities. Prior research has mainly evaluated ethical outcomes after deployment, yet their foundations are laid earlier, during development. At this stage, developers must rely on deontological reasoning, evaluating design choices against moral imperatives such as fairness and non-discrimination, because actual consequences cannot yet be observed. In practice, deontological reasoning often falls short, as it is cognitively demanding and inconsistently motivated. Drawing on deontological ethics, we examine how perceived algorithmic accountability orients developers towards deontological judgements and behaviours, and the conditions under which this effect is amplified. Using a multi-method design comprising a survey (n = 148), an online experiment (n = 102), and qualitative interviews (n = 15) with ADMS developers, we find that developers perceiving higher algorithmic accountability engage more in deontological judgements and behaviours. These relationships strengthen with greater ADMS knowledge and within a deontological work climate. We thereby move beyond high-level ethical principles for responsible ADMS design and demonstrate how organisations can elicit ethical behaviour from developers through algorithmic accountability.