Exploring Algorithmic Fairness and Trust in Generative Artificial Intelligence

GenAI has been used to personalise more contextualised recommendations compared to traditional AI. Personalisation in GenAI reflects the system’s ability to recognise each user as a unique entity through adaptation of content to users’ preferences and past behaviour (Rhee & Choi, 2020). To date, individual responses to algorithmic recommendations by GenAI remain divided (Shin et al., 2025). Given that many GenAI algorithms are primarily trained on English-language and Western-centric databases and are predominantly designed by White males (Budhwar et al., 2023), the datasets used in algorithms reflect long-standing structural inequalities embedded in systemic biases that may exacerbate existing inequalities and differences (Akter et al., 2021). Therefore, a focus on diverse ethnic groups is crucial in order to unpack the limitations and potential of GenAI innovation (Cabiddu et al., 2022) to promote inclusive marketing and collaborative human-AI models (Shultz et al., 2022; Pizzi et al., 2025).

Guided by social identity theory, a quasi-experimental study was conducted on 340 users to explore the multigroup effects on the relationships by comparing ethnic group differences (non-minority versus minority). Interestingly, the results indicate that the ethnic minority group is more concerned about how the personalisation directly works to influence their perceived algorithmic fairness and trust compared to the non-minority group when obtaining recommendations from GenAI. Finally, a qualitative study with 14 semi-structured interviews was conducted to reveal nuances in the relationships identified in the previous study. The qualitative results corroborate the findings from the quantitative study, revealing that ethnic minorities often perceive themselves as less relevant by GenAI due to the concerns that default datasets used in algorithms are from Western-centric and English-based datasets.

These findings offer significant theoretical and practical insights into how GenAI can be more effective in delivering social sustainability through capturing diverse expectations and preferences.