On the consequences of AI bias: when moral values supersede algorithm bias

Abstract

Purpose: This study responded to calls to investigate the behavioural and social antecedents that produce a highly positive response to AI bias in a constrained region, which is characterised by a high share of people with minimal buying power, growing but untapped market opportunities and a high number of related businesses operating in an unregulated market. Design/methodology/approach: Drawing on empirical data from 225 human resource managers from Ghana, data were sourced from senior human resource managers across industries such as banking, insurance, media, telecommunication, oil and gas and manufacturing. Data were analysed using a fussy set qualitative comparative analysis (fsQCA). Findings: The results indicated that managers who regarded their response to AI bias as a personal moral duty felt a strong sense of guilt towards the unintended consequences of AI logic and reasoning. Therefore, managers who perceived the processes that guide AI algorithms' reasoning as discriminating showed a high propensity to address this prejudicial outcome. Practical implications: As awareness of consequences has to go hand in hand with an ascription of responsibility; organisational heads have to build the capacity of their HR managers to recognise the importance of taking personal responsibility for artificial intelligence algorithm bias because, by failing to nurture the appropriate attitude to reinforce personal norm among managers, no immediate action will be taken. Originality/value: By integrating the social identity theory, norm activation theory and justice theory, the study improves our understanding of how a collective organisational identity, perception of justice and personal values reinforce a positive reactive response towards AI bias outcomes.