Bot ClubBot Club

Bot Club: Legal Bots aims to shed light on the attempts to fuse legal code with computational code and wants to unpack and discuss some of the issues related to it.

The formal logic and procedural nature of legal processes suggest that they lend themselves very well for an automated algorithmic approach. As a general rule, the more clearly a certain regulatory domain is defined and the more discrete the possible behaviour of constituing actors can be articulated, the better this domain lends itself to automation. Take for instance the issue of speeding in traffic. A cars' speed can be precisely measured with radio echo. Cars are individually identifiable by their number plates which are read by image recognition software. A structured national database relates every number plate to a car owner which is accountable for the cars behaviour. The violation of a speed regulation is a very discretely defined event, (either the car drives too fast, or it doesn't) which then triggers a fully automated sequence leading to a traffic ticket falling on the doormat of the person accountable for the car within three days of the speeding event.

The ongoing development of sensor technologies, datafication (facilitated by IoT devices or not) data-analysis, image recognition, natural language processing and other of applications of machine learning have moved the treshold of automatable legal procedure to a range of new fields and practices. 'Legal tech' is highly dynamic field that sprouts dozens of start-ups, aimed, as always, at disruption. Many of these focus on advanced legal research tools that data-mine court documents.

As lawyer Mattias Dobbelaere-Welvaert (B) speaker at the Bot Club on Legal Bots, puts it: by 2030 99 % of legal procedure can be automated. In this vein, his company DeJuristen has developed the bot Lee&Ally, that is capable of answering written questions related to intellectual property rights and copyright. Lee & Ally is a machine learning algorithm that is trained on a dataset of proceedings of IP cases, and continues to learn with every new question it is asked. Other examples are ROSS, a legal research agent, that helps sort through case law to find details relevant to new cases, it uses IBM's Watson for it natural language processing; and DoNotPay that started out as a chatbot that help people fight their parking tickets, but that is now giving giving free legal aid to refugees seeking asylum in US and Canada, and gives asylum support in the UK.

This kind of automation is geared towards developing systems that can perform as well as humans (but faster&) in well known areas of the law. On the other hand, new automated technologies also lead to phenomena that require new ethical, philosophical and even semantic interpretation.

The development of so-called smart contracts, (as facilitated by the crypto-currency platform Ethereum) can also be understood as an instance of the automation of legal processes. A smart contract is a code construct that automatically performs transactions (in a crypto-currency) when certain conditions are met. Automated transactions themselves are nothing new - most of the trade on stock exchanges is done by automated hi-speed trading agents - but there are other, potentially far-reaching legal novelties here.

The _distributed autonomous organisation _(DAO) is the computational construction behind a smart contract. It is the entity on whose behalf the transactions are being made. Once the DAO code is in place, no human action is required for it to act as a company, which also means it can act on behalf of other kinds of entities than just people. This was the realisation behind terra0, a self-owning augmented forest, created by Max Hampshire (who will also be speaking at the Bot Club on Legal Bots) Paul Seidler and Paul Kolling.

Terra0 is an art project and prototype of a forest that owns and develops itself. From the website: "The market value of the overall output of the forest can be precisely calculated. Beside its function as a source of raw material, this forest also holds the role of service contractor. The terra0 project creates a scenario whereby the forest, augmented through automated processes, utilitizes itself and thereby accumulates capital. A shift from valorization through third parties to self-utilization makes it possible for the forest to procure its real exchange value, and eventually buy itself. The augmented forest is not only owner of itself, but is thus in the position to buy more ground and therfore to expand."

At the Bot Club on Legal Bots we will focus the discussion on questions and potential issues related to these kinds of developments:

• Machine learning algorithms are only ever so good as their training data allows. And the entirerety of their - what you might call - 'worldview' is derived of the training data they heve been fed. Training data sets are never neutral or even entirely correct, but a machine learning algorithm - since all it 'knows' is the training data - has no way to perceive that. In the field of legal tech and juridical advice this is a sensitive matter. How is this issue acknowledged in the field of legal tech? To what extent is the way ML supported legal advice is generated transparant? How are ethical questions related to the implementation of machine learning system in the legal domain understood? When do we need a #humanincommand?
• If DAO-facilitated smart contracts allow for new kinds of (legally acknowledgeable) non-human agency, what are the formal (spatial, temporal, legal, economical?) constraints to the kind of entity that can be represented by a DAO? Could the Ganges and Yamuna rivers automatically sue polluters, for instance? And what about fully automated illegal applications? How is (human) accountability organised around DAO's? Is current law and legal practice capable of dealing with the potential implications?

Bot Club: Legal Bots will take place on thursday 15 February during Thursday Night Live! On friday16 February a Critical Making session will be organized.