How Emotionally Intelligent AI is Changing eDiscovery
There’s no denying that Machine Learning and Artificial Intelligence (AI) are transforming every facet of our lives. Now, a new advancement in AI technology that detects emotion in documents, emails and chats is finding itself into eDiscovery systems. We call this, emotionally intelligent AI.
We regularly speak with automated customer service representatives and see ads on our Facebook feeds that were chosen specifically for us by AI-powered technology. The eDiscovery industry is being impacted by this new technology as well. Machine learning and AI are becoming crucial to the discovery process, quickly sifting through massive amounts of data so that legal teams can locate relevant information and make quick and accurate review calls.
Despite these important advancements, we would be naive to think that we’ve seen the pinnacle of AI’s heyday. As the technology grows in complexity, it’s becoming more intuitive and perceptive, and it has a greater ability to pick up on subtleties that might have eluded it before. One new iteration of the technology, emotional intelligence, is already showing the potential to change eDiscovery as we know it.
What is emotional Artificial Intelligence?
Emotionally intelligent AI examines data with a higher level of comprehension than what technology has been able to do previously. Unlike an actual human interaction, with written documents, chats, emails and the like, there is no face-to-face information to draw from, such as body language or voice inflection. However, AI is able to pick up on other contextual clues — such as the circumstances around the data’s creation (who created it and when, for example), the speaker’s probable intent, the author’s emotional state and even notable events occurring at the same time as the conversation. With these indications, the information is transformed from bytes of data into a human story.
Traditionally, eDiscovery has relied on date range and keyword-based techniques to cull through hefty amounts of data. These include Boolean searches, which incorporate modifiers like AND, OR or BUT, and proximity searches, which produce results based on the closeness of search terms to one another. Though certainly valuable, these tools are unable to filter down data beyond a certain point. Likewise, the most commonly used Artificial Intelligence, Technology Assisted Review (TAR) analytic tools, can only make determinations based on bland content alone. Emotional intelligence, on the other hand, helps machines identify the nuances and intentions behind those interactions, often revealing important clues and colors that can dramatically change one’s understanding of that data.
How does emotionally intelligent AI work?
With the addition of emotional intelligence, AI examines data for context and significance instead of just viewing words as a series of letters with no real meaning. For example, it takes into account a range of situations and emotional states, including positivity or negativity, opportunity, intent, pressure, and rationalization. The words themselves are a factor — derogatory words, for example, are a clear sign of negativity. Punctuation and sentence phrasing are also considered, such as the frequency of exclamation points or capital letters. These indicators reveal the new meanings that the technology picks upon. Even without offensive language or other clearly negative words, a document could still be flagged as negative due to other determinants.
The beauty of this artificial intelligence technology is that it does not require the training of a system. Emotional aspects can be factored in from the beginning of the eDiscovery process, and ultimately, this gives attorneys more accurate insight into data, eliminates irrelevant files earlier in the process and allows the time spent on human review to be hyper-focused on the documents that matter most.
In addition to the context and intent of words and phrases, the technology factors in abnormal behavior, such as frequent communication outside of business hours or unusually regular conversation between two people during a certain timeframe. Emotionally intelligent AI considers the subject under discussion during those conversations, whether that topic is typical for those involved and the possibility that certain words were chosen to disguise the true nature of the conversation. The technology even considers who is talking to whom and if relevant events occurred during the same time frame.
The information discovered as a result of emotionally intelligent AI can offer an entirely new angle to the data, and documents that might have taken a significant time (and money) to find before can potentially rise to the top of the pile with just a few clicks.
AI using emotions
Let’s look at a real-world example of emotionally intelligent AI in action: A condo association suspected that its controller and another person had been siphoning funds for many years. It hired a forensic investigator who collected between three and four terabytes of data from four custodians, including the two suspects. First, the investigator searched the data for the two suspects’ names, as well as keywords they may have been using. This reduced the population to a few hundred thousand documents. Then, he searched the remaining data using emotional descriptors — specifically, negativity and opportunism. Adding this emotional layer to the keyword strategy produced a near-exact set of documents the investigator was seeking — and did so within seconds. The resulting information is currently being used for legal action by the condo association’s members.
In addition to corporate internal investigations like that, emotionally intelligent AI can be used in sexual harassment cases, product liability cases and others. The technology picks up on the context behind conversations, such as when a person is being pressured to do something they shouldn’t or has an opportunity to embezzle or commit fraud.
It’s even possible that the technology could one day be incorporated into the workplace communication system — such as a plug-in to Microsoft 365 — allowing a corporation to monitor employee emails and chats for certain components, such as negative intent or harassment. Once triggered, the system would automatically notify management, who could review the data and determine if there is a valid reason for concern.
Even though this integration hasn’t happened yet, emotionally intelligent AI has already changed the scope of eDiscovery, and we expect it to become even more useful as it continues to advance. The incorporation of emotional context into search results allows the technology to draw precise and intuitive conclusions, ultimately saving legal teams time and money during the review process.