The problematic persistence of algorithmic bias is a mystery to me. Not, as it happens, because I am particularly baffled or confronted by this form of technological encoding and mediated emergence of core social and cultural inaccuracies, assumptions and deep-seated human insecurities. It is much simpler than that, in the end.
Machine-intelligence powered classificatory systems are implicitly biased along all the faultlines and socioeconomic or ethnic, class-based and racial stereotypes and generalisations that we and the complex contexts we inhabit are also prone to, if for many of us at an unconscious level. Machine learning used to facilitate bank loan risk assessments, court sentencing regimes or criminal and gang profiling have been shown time and again to replicate and reinforce the inequitable and often enough dramatically unjust biases and cultural or (ideological) generalisations that – for better and for worse – define the shared consciousness of a time and place.
What puzzles me is not the emergence of these biases – as such complex self-expression and hyper-extended cognition of technological artefacts is always and to some extent only ever going to hold a mirror up to ourselves. The complete mystery here is that, even while our darker natures and dissociative cultural neuroses are reflected back to us with such stark algorithmic clarity, the primary focus of the developers and the runaway train of commercial hype that surrounds artificial intelligence has been to attempt to (merely) brute force the algorithms themselves into submission, to cause these partially-autonomous measures of our dissonant selves to obscure the unsavoury facts that they have discovered from within the multiplicity of data points these machines were trained on.
Rather than attempt to remediate a broken psychological and cultural (and as often – political) system, the game here becomes one of obscuring the problems that these autonomous systems have underlined and dramatically outed. Finding that some of our most sophisticated technologies reveal that there is a bitter and ugly streak running through our collective heart, the immediate institutional and commercial, corporate response is “how do we fix the technology?”, not “how do we fix ourselves?”.