This research project takes a specific position regarding the relationship between law, lawyers and legal technology. Building on the philosophy of technology we emphasize that tools often transform their users as well as the kind of society they build and sustain.

This has normative implications, because though technology in itself is neither good nor bad, it is never neutral. Focusing on the normative implications for the law, the project asserts that text-driven law affords a different normativity than data-driven law, for instance because machine experience (data) is different from human experience (legal text, human action). The same goes for the affordances of code-driven law, which affords a specific type of enforcement (as it is self-executing) and may even be said to pre-empt the need for enforcement.

Normativity, in this context, should not be equated with morality. Normativity refers to the mutual expectations that enable meaningful action, whereas morality involves the moral evaluation of such expectations. Normativity also regards their force (inducing or enforcing, inhibiting or ruling out), their scope (targeting a specific or generic range of actions), their scale (affecting face-to-face or much larger populations) and their transparency (text externalises norms, whereas as architecture or nudging may entail normative effects without such norms being visible and thus contestable).

Text-driven normativity is deeply entwined with what speech-act theory has coined the performative nature of speech acts that do what they say: declaring two people man and wife actually attributes all the legal effects of marriage, instead of merely describing the marriage. The normative effects of text include the uncertainty that comes with the ambiguity of natural language, which relates to issues of interpretability that are in turn related to the contestability of norms that have been externalised (i.e. publicised). For an in-depth inquiry into text-driven normativity and legal protection, see the first Working Paper on the subject and Hildebrandt’s article on The Adaptive Nature of Text-Driven Law.

Data-driven normativity is deeply entwined with the feedback loops generated by predictions (cybernetics): if ‘the data’ predict that one is inclined to criminal behaviour, the police, the psychiatrist and even the subject themselves will base some of their decisions – in part or entirely – on such predictions (whether or not they are correct; noting that this is hard to test outside the laboratory). To the extent that the subject themselves are not aware of the predictions, decisions may have subliminal effects, noting that the underlying techniques, such as machine learning, entail a number of specific interpretability problems. This could be summarised by suggesting that data-driven normativity is in many ways pre-emptive (aligned with nudge theory).

Code-driven normativity is deeply entwined with the automation of compliance with pre-set standards: if certain conditions are met the code will self-execute whatever it was programmed to ‘do’. The force of code-driven normativity is in many ways top-down, not leaving room for disagreement about the right way to interpret the norms. The translation of the norm into code implies an effective choice for one particular interpretation, that cannot be challenged without huge effort (if at all).