Brief History of LegalTech
Until the late 1950s, the use of information technology in law—especially regarding substantive law solutions—can be broadly described as an evolutionary process from deductive to data driven approaches.
The deductive approach (also called rules-based logic) expresses information processing as a logical, step-by-step procedure. Example: If a certain condition X is fulfilled, Y needs to be verified. In the early days of LegalTech, it was assumed that most tasks could be described in deductive instructions by attempting to replicate computer-based versions of human cognitive processes (Stuart Russel/Peter Norvig, AI: A modern approach, 2010, 3-10). Behind this focus was a belief that computers have to “work” the same way as the human brain. The attempt was to create an artificial version of the human brain functionality. As the deductive approach did not show the envisaged outcome, the data driven approach became (until today) increasingly popular and more promising. The data driven approach is used to describe information processing that cannot be articulated as a mere series of logical steps. These inductive rules take the form of statistical equations that model the relationship between the information inputs and the processed output. The equations are estimated or trained with samples of historical cases. The estimated equations are then used to process new cases.In the mid-1980s, legal expert systems—being a deductive approach—were en vogue. A legal expert system is a computer system that emulates the decisionmaking ability of a human expert in the field of law. Legal expert systems employ a rule/knowledge base and an inference engine to accumulate reference and produce expert knowledge. Expert system builders gather expertise and knowledge from (legal) experts. They then incorporate the knowledge and expertise into a software application in order to make the expertise and knowledge easily replicable.
Some legal expert systems, like Susskind and Capper's, The Latent Damage System, became publicly accessible (cf. Philip Leith, The Rise and Fall of the Legal Expert System, EJLT 2010, 1 ff.). Many saw legal expert systems as an opportunity to provide inexpensive and easily accessible legal knowledge and advice. Expert systems are (early) forms of AI software.In the 1980s and later, soft computing, a combination of technologies such as fuzzy logic, probabilistic reasoning and neural networks, became fashionable without sustained success. In the 1990s and beyond, expert systems were considered a failure because they only seem to work if (a) the legal rules are straightforward enough, (b) there is no ambiguity or vagueness regarding the inputs, and (c) there is clarity about which rule applies in each situation. Expert systems could not account for some of the most important performances of legal cognition: making analogies, applying vague and imprecise standards, using insufficient or contradictory data, learning through examples and experiences. What followed was the so-called AI winter of the late 1980s, a period of reduced funding and interest in legal AI research. The initial hype was followed by disappointment, criticism and funding cuts. Today, legal expert systems (sometimes also referred to as decision support systems) still play an important role by standardizing, formalizing and modelling expert knowledge. Nowadays, however, the aim is not to eliminate the human factor but to build expert systems that augment and facilitate the decision making process of lawyers.
Another trend in the 1990s was the development of formalisations of domain conceptualisations (so-called ontologies). A legal ontology tries to indicate concepts that exist within the legal domain and seek to explain how these concepts are related to one another. In other words, legal ontologies are used to express the common understanding of concepts and relations among them in a formal and structured manner.
Ontologies can be used to, for example, create a common gateway facilitating exchange between domains such as taxes and property administering. Ontologies can be also very useful for information retrieval as it is possible to detect similar things that are named or called something else. Early examples include Valente's functional ontology and the frame based ontologies of Visser and van Kralingen (both 1995). There are many examples ranging from generic top-level and core ontologies to very specific models of particular pieces of legislation.Even though not all early approaches to enhance legal work with technology were equally successful, they all contributed some important spadework for current LegalTech tools. They especially lead to the conclusion that it is not promising to only re-model an “intelligent” process that replicates the thinking of the human specialists to achieve “intelligent” outcomes. Today's LegalTech focuses on whether a computer system is able to produce accurate, appropriate, helpful and useful results which could be considered as “intelligent”. In the following, we will explain which “modern” technologies are used nowadays to imitate intelligent processes.
3