From Distrust to Trust : Regulating AI - Based Regtech

Scholars and practitioners have long recognised that trusting artificial intelligence is fundamental to its use, adoption and integration into practice. However, trusting a technology that is fraught with concerns over functional opacity and potential bias can be largely problematic. Flowing from this, the dominant approach towards building trust in AI has been through the development of Trustworthy AI. The paradigm of Trustworthy AI is underpinned by conventional notions of trust and trustworthiness (Hawley, 2017 ; Sitkin and Roth, 1993), marked by a belief that ensuring trustworthiness in AI will lead to trust in the system. This paper challenges this conventional and linear understanding of trust and the relationship between trust and trustworthiness. More particularly, in this paper I challenge the understanding and assumption that trust will be causally motivated by the trustworthiness of the trustee. Trust is a multidimensional and complex cognitive process. Trust often arises from subjective human perceptions that are only remotely related to the trustworthiness of the trustee. Using the case of AI -Based Regulatory Technology (Regtech), I examine how trust in AI system remains elusive and how distrust develops despite measures aimed at developing Trustworthy AI. In this paper, I argue that regulation, in particular, polycentric and responsive regulation (Ayres and Braithwaite, 1992 ; Black, 2003 ; Scott, 2001) has a central role to play in addressing the distrust of AI-Based Regtech , thus in the process, creating and maintaining trust. I further argue, that it is only in presence of such a regulatory framework that Trustworthy AI systems (in this case, trustworthiness of AI-Based Regtech) could effectively contribute towards building trust in the technology, as opposed to trustworthiness building trust by default.