Training an AI Priest (in collaboration with Dane Watkins)

As discussed in my previous blog post,  I problematise the idea of  datafication as a process that turns research subjects into square data points/units of observation, often limiting life chances and capacity  for self-determination (such as in the cases of the education, health, social, legal or commercial systems).

Thus, as part of this fellowship, I am investigating  ways to reassert the voices, meaning-making abilities and agency of research subjects, and  raise critical awareness of the bias involved in the datafication process.  For this, I am exploring – in collaboration with Dane Watkins – the use of different carnivalesque practices, (such as amplification, absurdity, bricolage, reflectivity and  grotesque reversal) that could help reassert the voices and subjectivities within systems and deflect scrutiny from the object of research onto the process of datafication itself.

The confessional is an experimental project that aims to draw attention to and amplify  bias involved in the datafication process.

The use of the priest is a nod to Dane’s artistic and academic work on the world of medieval scribes and artists, and their emphasis on the humorous, the monstrous and the grotesque  within religious texts…For centuries, religious writers pondered the right way to  devise penance according to the nature, severity of the sin and the identity of the offender. Thus, this is also enabling us to make a parallel with the judiciary system currently trialling AI to determine conviction and penance. In a loose comparison, analogue algorithms were used by priests in medieval times to judge sins and impose penance, much like we tentatively use digital algorithms to emit judgment on crimes and devise penal sentences.

Can we imagine a priest that blurs the line between mysticism and science, using a computer algorithm to judge sins and determine penance? This is what inspired our carnivalesque confessional project where you, as part of our virtual congregation, are  invited to confess your fictional or actual sins to Father Ludo Vicarious, our virtual AI priest.

The project is divided into 2-phases, with an emergent design that allows us to learn and adjust our approach in the process. In the  first phase, we developed a user interface, linked to a database that uses anonymously crowdsourced confessions to train the AI priest in identifying and quantifying sins.  The second phase of the project will feature our actual AI Priest, greeting confessors and doing his thing: listening to sins, analysing them, quantifying them, providing an ‘adequate’ medieval penance for absolution, and perhaps, recommending other sins of interest…

Truth and veracity is not the point here, quite the opposite… it is rather about reinstating disbelief in the datafication system and drawing attention to the possibility of error, to the amplification of bias. The project is carnivalesque in its use of amplification…. or rather, by demonstrating that the algorithm itself is a source of grotesque amplification of bias. Is Father Ludo Vicarious that far-fetched from our current use of AI judgement in the judiciary system?

Grotesque is further encoded in the style, as well as the tone that emphasises the jestering nature of the project… defeating any  pretension to  an actual quantification of sins, and by extension, wholesale quantification of human existence.

The idea is particularly potent as it pokes fun at Christianity – long accepted as a dominant truth paradigm and now more openly challenged in Western liberal circles… ironically, given our standing devotion for positivism and datafication.

The confessional,  whilst enabling parallels and questions about mind/body control, authority, intrusion, sacrilege and bias, questions this propensity to take data too seriously and uncritically (as something verging on the divine).  Will we, one day, be able to poke fun at, and take a reflective distance from our drive to encapsulate the world into/as data?