Artificial intelligence in law has witnessed huge changes over the years. It's grown into a solution for many of the day-to-day tasks legal professionals face and is used to combat long working hours by offering significant time savings.
Throughout this blog post, we'll explore a short history of artificial intelligence, analyse its impact on the legal sector and demonstrate how it assists with legal reviews.
A brief history of artificial intelligence
The formal discipline of artificial intelligence began in the mid-20th century, with the term being coined in 1956 at the Dartmouth Conference.
There was much optimism in the early years of AI and researchers envisioned the development of machines capable of learning, reasoning and problem-solving with minimal human input.
In the 1960s and 1970s, research focused on symbolic AI, using logical rules and symbols to represent knowledge. Progress was slow, and the 1980s experienced an 'AI winter' as funding and interest diminished.
The resurgence of AI began in the 1990s with machine learning approaches, with breakthroughs in algorithms and computing power facilitating significant progress in the late 20th and early 21st century.
In recent years, AI has become part of everyday life. It's integrated into many industries, from autonomous vehicles to healthcare, entertainment and law.
Artificial intelligence in law
Much like many other industries, artificial intelligence has been transformative to litigation. Integrating AI tools has allowed firms to streamline processes, enhance efficiency and provide valuable insights.
Let's look at some tools legal professionals can use to help their processes.
Machine learning
Within artificial intelligence in law, machine learning allows computers to learn and progressively improve performance on specific tasks without the requirement for programming.
Machine learning can be divided into three categories: supervised, unsupervised and reinforcement.
Supervised learning
In supervised learning, the computer is taught using a training set that provides inputs and desired outcomes. The computer learns a general rule, which it applies to further data.
Unsupervised learning
The computer is not fed any training set but learns from decisions made or finds structure in the input data. This is often the purpose of unsupervised learning: to uncover hidden patterns in the data.
Reinforcement learning
This is where the computer interacts with a dynamic environment where it performs tasks and receives 'rewards' for completing them. The programme aims to maximise these awards to increase learning.
When we dig deeper into supervised learning, we then find active learning. Active learning is a special case of machine learning in which the learning algorithm can interact with an existing data source to label new data points with desired outcomes.
Active learning in law
Over the past 20 years, AI and machine learning algorithms have replaced many mundane, day-to-day tasks and processes within the legal industry. AI is now commonly used in all levels of law, from aiding judges in making prosecution decisions to helping businesses manage GDPR and compliance issues.
However, the area in which AI is pushed towards its maximum is within eDiscovery in Technology Assisted Review (TAR).
TAR is the process of using technology to support reviewers throughout the document review process. Over time, there have been several iterations of TAR as active learning has advanced, and although many now refer to TAR as continuous active learning, this hasn't always been the case.
Predictive coding, now commonly referred to as TAR 1.0, is the predecessor of active learning. It's the process of training an algorithm using a training set of data that it can learn from and replicate on another data source.
The algorithm is given a base set of criteria that relevant documents meet and proceeds to filter out documents from your review pile that don't meet this criterion, reducing the number for review.
Active learning is often used ahead of predictive coding as it provides more relevant information. The main difference between the two is that, where predictive coding identifies just relevant and non-relevant documents, active learning identifies the documents and presents them to you in a prioritised view. Therefore, you see the most relevant documents first.
Active learning in RelativityOne
Continuous active learning (CAL) is used within RelativityOne software to drastically reduce the time and money spent reviewing cases that suit the algorithm's capabilities.
CAL doesn't adapt to every circumstance and there are certain criteria a case must meet for it to be successful.
- Your case should have a large number of documents of a similar type.
- Your case shouldn't have too many image files, as the algorithm searches for keywords and phrases that reviewers have deemed relevant.
- Your case documents should all be in the same language. If your case contains multiple languages, it's best to segment the data into those languages and conduct separate reviews.
The accuracy of active learning
Active learning isn't completely error-proof and is only as good as the material it's learning from. If you code the algorithm with a 10% sample set with lots of similar data, the algorithm will work well to identify data with common patterns across the rest of the review pile.
However, if relevant data of a different type is within the review pile that wasn't part of the training sample, the algorithm won't know to code it as relevant. This is where the ability to check the accuracy of your algorithm becomes extremely useful and where the elusion test comes in.
An elusion test checks the accuracy of your active learning algorithm by providing a random sample of documents coded as 'not relevant'. You can then review these documents for relevance, and any that are deemed relevant are seen as 'missed' by the algorithm.
At the end of the test, the number of missed documents from the sample set is calculated to estimate the number of missed documents in the discard pile. From this, your team can decide whether to review the remainder of the documents or end the review.
To conclude
Active learning and AI have massively changed the landscape for legal technology and the industry as a whole. However, it's vital to remember that without human input to oversee, correct and provide the basis from which algorithms learn, these technologies would have little use.
Many worry about artificial intelligence's impact on law and other industries, but humans are far from being replaceable.
All in all, artificial intelligence can benefit legal professionals when used correctly. Time and money savings can be made, making learning about available programmes and solutions extremely worthwhile.
Learn more about artificial intelligence in law
If you want to learn more about artificial intelligence in law, our Content Hub is the perfect place to start.
Packed with expert insights, our Content Hub gives you free, unlimited access to our range of guides covering various topics.
To access the Hub, simply click below.
