In our the first article in the series, we discussed the intricacies of the intelligent cull – a vital tool in combatting the ever-increasing data volumes involved in eDiscovery. Today, we’re taking a look at the other side of that coin – using AI and machine learning to identify what to include in your dataset for investigation and, potentially, review.
As always, the stakes are high. Include too much data and review costs will be astronomical. Include too little, on the other hand, and a critical piece of evidence may be missed.
The trouble with “fact patterns”
By the time a matter reaches the legal department or law firm, there is generally a suspicion of a particular wrongdoing. This suspicion forms the basis of what we like to call the fact patterns – the persons of interest, date ranges and traditional disclosed keywords and phrases that are used to focus the investigation of the alleged misdemeanour and reduce datasets to manageable volumes.
The trouble is: fact patterns don’t come with any guarantees of correctness or completeness. There is every chance that the original fact patterns only represent a fraction of a much larger puzzle. This is particularly true – and challenging – if the matter is still early in the investigatory phase.
Dealing with unknown unknowns
So how do you find what you need when you don’t necessarily know what you’re looking for? As it turns out, the trick is simply to initially cast the net a little wider.
With the help of AI and an investigative mindset, it’s possible to automatically expose leading indicators from within a much larger dataset. In other words, entities (people, places and companies), topics and aligned concepts, patterns of behaviour, sentiment in communications, and even unseen communicators, all with demonstrable connections to the matter at hand.
These indicators (all examples of readily consumable, graphically presented cognitive analytics capabilities) can then be used to validate, correct, and/or expand the original fact patterns to more accurately ascertain the true shape and dimensions of a matter.
AI’s behavioural and emotional edge
AI capabilities have evolved in leaps and bounds. Where previous analytics would have found only text, today’s AI (when correctly trained and deployed by a skilled user) can bring context and hence improve the understanding of that text and conceptualise its behavioural and emotional implication.
When combined with pattern analysis, this enables AI to recognise specific tones and inflections for each individual. This can then be used to recommend and prioritise documents that may otherwise escape scrutiny, but have a high likelihood of relevance to the case.
Reveal’s AI offering
Reveal offers a number of AI and ML tools and capabilities to support more accurate and comprehensive early case assessment. These include:
-
- An AI model library and marketplace, where investigators can source highly-specific, out-of-the-box, pretrained, machine learning AI models. Each has been algorithmically developed by Reveal to meet the real needs of different audiences, practice groups and teams.
-
- Linguistic intelligence leveraging natural language processing to explore and uncover hidden connections between people, places and things.
-
- Emotional intelligence able to automatically recognise positive or negative discussions around an event and assign a sentiment score which can be used to prioritise evidentiary “hotspots”.
-
- Behavioural intelligence that combines linguistic intelligence, content classification and emotional intelligence to piece together stories and events from patterns and connections found within the data.
-
- Continuous active learning that automatically incorporates each decision made by subject matter experts to improve the efficacy of the model and more accurately prioritise document recommendations.
It’s also worth noting that Reveal’s AI model library grows continuously as new needs are identified. However, organisations can also develop and securely store their own versions of models, trained on specific topics or using industry-relevant language, to be reused on similar cases down the line.
Needless to say, this can dramatically accelerate early case assessment on these matters, contributing to better decision-making, and quite possibly reducing overall eDiscovery costs.
Conclusion
Deploying AI and analytical tools earlier in the eDiscovery process can vastly improve early case assessment by providing more accurate and complete fact patterns, surfacing hidden connections and evidence trails, and prioritising the most relevant documents for early review.
The result is a far more comprehensive early understanding of the matter at hand, enabling more confident decision making around case viability and direction, expediting progress, reducing risk, and potentially cutting costs.
Sound interesting? Get in touch to find out how we can solve your eDiscovery challenges.
Read more of our series on practical AI for eDiscovery
Practical AI for eDiscovery: today, tomorrow and in future
We’re still a long way away from discovering where artificial intelligence will lead us. But preparing for that mysterious future shouldn’t stop us from making the most of what we have here and now.
1. Intelligent culling - a critical component of cost-effective eDiscovery
The most effective way to reduce data volumes for review and hosting is by using AI to intelligently cull irrelevant, duplicate and otherwise non-responsive data.
2. Using AI to improve inclusion
How do you find what you need when you don’t necessarily know what you’re looking for? With the help of AI and an investigative mindset, it’s possible to automatically expose leading indicators from within a much larger dataset.
3. Expectation vs reality: what Generative AI really offers eDiscovery
Is generative AI really the next frontier in eDiscovery? How much of the hype is grounded in reality? We explore practical applications for GenAI in eDiscovery.
4. Finetuning Generative AI for eDiscovery
AI may be powerful, but it still requires human input to deliver high quality results. We share our GenAI learnings around how context and prompt influence output and how GenAI output can be finetuned.