Fraud and deception detection: Five language fingerprints

Last month I described how computer-aided text evaluation may also help uncover fraud and deception in corporate communications, but what other insights can we gain from this investigation of scandal-ridden corporations?

We used Deception and Truth Analysis (DATA) We examined ten of the most important corporate scandals in recent history and located that the typical lead time between our textual identification of the fraud and public awareness of the potential scandal was greater than six years.

Corporate scandals: time between written evidence and public recognition

The obvious query is why. Why does it take so long for regulators and markets to discover these scandals? And a follow-up query: what insights from text-based analytics can we use to discover these scandals earlier? Let’s undergo them one after the other.

Theory: It is the behavior

Why does DATA detect deception faster than interested investors and regulators? After excited about it for some time, we developed a theory that boils all the way down to 86.5%. That is the proportion of economic information that’s expressed in text somewhat than numbers in annual reports. Text communication reveals the behavior of corporate management teams, and that behavior results in the result that’s expressed in numbers.

So 6.6 years between the primary indication of fraud and the exposure of the scandal is the typical length of time that a misbehaving company can fake its fraud until it simply can not manipulate its figures.

What’s interesting is that the 2 scandals that got here to light greater than a decade later each involved financial firms: AIG and Lehman Brothers. Their annual reports were tons of of pages long, and the speed of cash through their balance sheets and profit and loss statements was very, very high. So it took a substantial period of time for his or her bad behavior and decisions – the inputs – to finally show up within the numbers or the outputs.

If this theory provides a sound explanation for this lead time, then there ought to be linguistic fingerprints on scandals that investors can search for either as an early warning system or as a second opinion to the conventional fundamental work that investment research teams do.

Language that reveals a possible scandal

After examining the ten scandals mentioned above, in addition to Wirecard and other recent controversies, we identified five text fingerprints that differ by greater than 50% from those of more honest corporations.

Scandalous words and company communication

In addition to text-based evaluation, we also conducted one-on-one interviews to raised distinguish between deception and truth and to discover some of individuals’s more pan-cultural deceptive behaviors. Our results were consistent with what previous researchers in the sector of lie detection had found: that every of the five potential deception indicators revealed by text-based evaluation also emerged in face-to-face interviews.

So let’s take a better have a look at each.

1. Words that express friendship

Researchers who detect lies have shown that fraudsters often use obfuscation to create confusion. One way they do that is by utilizing words that imply friendship in business communications more often than is typical. Fraudulent corporations use such terms 56.1% more often than average, our evaluation found. So if an annual report accommodates a lot of flattering terms, it could possibly be evidence of obfuscation and deception.

But a distinction is crucial here: words that indicate friendship – “friend”, “buddy”, “neighbour” and “gang”, for instance – are different from friendly words.

2. Risky words

Scandalous firms favor words that indicate risk to a much greater degree than the typical company. These include terms like “averse,” “avoid,” “concern,” “difficulty,” “prevent,” “stopped,” and so forth. These sorts of words are already getting securities analysts on edge, and as we discussed within the last article, corporations are proactively removing these sorts of “red flag” words from their annual reports.

3. Impersonal pronouns

“Another,” “everyone,” “somebody,” and “anyone else” are the sorts of impersonal pronouns that dishonest corporations use far more often—54.1% more—than their honest competitors. Why do they like to be impersonal of their communications? Researchers suspect that they try to create emotional distance between themselves and people they try to mislead.

4. Words that indicate a difference

Lying is cognitively demanding. One manifestation of that is that through the act of deception, the liar is commonly unable to tell apart between competing viewpoints in his communications and is due to this fact less inclined to make comparisons. So using words that suggest differences is definitely a cue to truthfulness. Constructions that present opposing viewpoints – “compared to other years…” – are examples of this.

Scammers even have an agenda: they need to persuade their victim to consider their preferred narrative. They are unlikely to tell apart between other narratives and usually tend to give attention to the one they like.

5. Words that negate an announcement

Research also shows that liars often use more negative expressions than individuals who tell the reality. That’s why we distinguished between words that express friendship and words which can be friendly.

But researchers don’t all the time conclude that cheaters are more negative than honest ones. However, our evaluation of dishonest corporate communications suggests that they use words like “not,” “never,” “shouldn’t,” “doesn’t,” and “mustn’t” at a rate 50.4% higher than average.

Bload

So what’s by far the strongest indicator of fraud? The variety of swear words in an annual report. Although they’re rare, swear words appear a whopping 277.1% more often within the annual reports of scandal-ridden corporations than the typical.

If you liked this post, don’t forget to subscribe.

Photo credit: ©Getty Images / Matthias Kulka