A key component of the forensic investigator toolkit: AI
5W1H Part 5: Why?
Today, we’re on to the why part of our 5W1H series exploring how eDiscovery tools and techniques can be used to answer the critical investigative questions of who, what, when, where, why and how (5W and 1H). In this article, we focus on the importance of AI as part of the forensic investigator’s toolkit to unearth sentiment and emotion in large volumes of data.
If you’re just joining us, click on the links below to read the earlier articles in the series. Otherwise, let’s dive into the ups and downs of using technology to shed light on the emotional intelligence aspect of eDiscovery.
AI’s role in uncovering why
Why can be one of the trickiest questions to answer in an investigation. It deals with “softer”, human concepts like motivation and rationalisation, and requires reading between the lines to analyse the context and sentiment behind events and actions.
Ironically, AI can be incredibly helpful in picking up contextual clues and emotional cues within content. This can be invaluable in unravelling the emotional threads surrounding a matter to reveal why something happened the way it did.
Why concepts beat keywords in an investigation
When it comes to answering why, investigators are unlikely to have a neat set of keywords to enter into a search query. As a result, capabilities like Reveal’s Concept Search are extremely useful.
Concept Search leverages artificial intelligence and machine learning to understand the context and meaning of the words within documents. This understanding is then used to suggest documents that are conceptually similar to the search query, enabling investigators to find relevant information even if specific keywords are not present.
The ten most likely related concepts returned are listed as Top Concepts. These can be assigned relative weights within the current context to refine the search for more documents and more relevant concepts. (Additional concepts suggested may be added to the list and weighted at investigators’ discretion.)
Having Sentiment Analysis in the forensic investigator’s toolkit
Of course, concept search can only go so far when it comes to answering why. To understand the drivers behind an event or action, you need to understand the sentiments surrounding it. In Reveal, this is achieved using natural language processing techniques like sentiment analysis to analyse emotional intelligence.
Sentiment analysis enables investigators to focus in on emotionally charged communications within the results of (for example) their concept search. It uses natural language analysis to identify the emotional cues within text, providing investigators with insight into the overall tone of a document as well as the emotional context of specific terms.
In Reveal, sentiment for each segment is also scored using an algorithm, making it easy for investigators to identify positive, negative and changing/flip-flopping attitudes to guide further investigation.
Sentiment Analysis and Fraud Triangle Theory
Fraud Triangle Theory, developed by criminologist Dr Donald Cressey, explains the three factors that lead individuals to commit fraud – namely: pressure, opportunity and rationalisation.
Understanding these elements can help organisations develop strategies to prevent and detect fraud by addressing the pressures individuals face, minimising opportunities for fraud through robust internal controls, and promoting ethical behaviour to reduce rationalizations.
Sentiment analysis can play a pivotal role in this process, revealing the pressures and rationales behind fraudulent behaviour. Analysing sentiment means diving into communications, however, requiring consolidated messaging and chat data as well as more formal email communications. These unstructured datasets can prove challenging without AI tools to sift and search through the millions of datapoints, zone in on communications of relevance using concept and keyword search, and then apply emotional intelligence to surface the most likely candidates for human review.
Rounding out the picture
Sometimes, fleshing out the picture of why means analysing more than just unstructured data. For example, at Salient, we’re often asked to use our forensic data analysis capabilities to perform transactional analyses on large volumes of bank statements looking for transactions that could shed light on the why of a situation and more.
We have developed our own bespoke visualisation tools and models to bring this data to life, including one that demonstrates relationships between entities using concentric rings. Working with large numbers of spreadsheets, these visualisations can bring together data like people, organisations, directors of companies etc. to highlight new lines of enquiry in relation to things like lifestyle audits.
Using tools and analytics to answer the
5Ws and 1H in a forensic investigation
1. Who
The best starting point for answering the question of “who” is involved in a matter usually centres on communications.
2. What
eDiscovery tools are invaluable to uncover the concepts, context and content under investigation.
3. When
Understanding when a digital footprint was left can be extremely valuable when aligning events in an investigation.
4. Where
“Where” is a multi-faceted concept that encompasses a variety of data types and requires the use of various eDiscovery tools and methodologies.
5. Why
Why can be one of the trickiest questions to answer in an investigation and requires reading between the lines to analyse context/sentiment.
6. How
Understanding how typically requires sifting through vast bodies of potential evidence to find the “trail of crumbs” and follow it to its conclusion.
We use our skills and advanced technology to help you find all the pieces in the puzzle faster and more efficiently. Find out how Salient Discovery could accelerate your next digital forensic investigation.