top of page

Sam Bankman-Fried's Language Foreshadowed the FTX Fraud


As the former CEO of FTX and “white knight” of cryptocurrency, Sam Bankman-Fried—also known as SBF—has been heavily scrutinized by the press since the collapse of his company this past November. In his latest interviews, SBF has presented himself in an innocent light while pleading ignorance of his misconduct (e.g., wire fraud, conspiracy to commit money laundering, embezzlement, and campaign finance violations), which ultimately led the U.S. attorney for the Southern District of New York to label FTX as "one of the biggest financial frauds in American history."


Since SBF’s arrest on December 13th, investors, lenders, and regulators have questioned whether there were warning signs that could have predicted his misconduct. SBF, and Bernie Madoff before him, used manipulation and deception to influence others. These types of individuals in power tend to possess “dark” personality traits, such as psychopathy. A lack of empathy, in particular, is cited as a hallmark of psychopathy on account of its associated social deficits, including a disregard for others and a propensity for criminal behaviour.


To gain a better understanding of SBF’s motives during his brief reign at FTX, we examined the degree to which he demonstrated facets of psychopathy. Using the transcripts of ten SBF interviews, conducted between 2020 to 2022 (six interviews before and four interviews after the collapse of FTX), we analyzed SBF’s language to look for evidence of linguistic markers of low empathy and social aversion. Our results illustrate the differences in SBF’s language during his interviews compared to the average CEO’s language obtained from a corpus of 4,218 CEO earnings calls of various companies.


Lack of Empathy


Research emphasizes the importance of assessing both emotional and cognitive aspects of empathy when assessing psychopathy. Emotional empathy revolves around feeling compassion for (rather than displaying callousness toward) others, whereas cognitive empathy revolves around understanding other people’s perspectives and emotions. Studies suggest that psychopaths tend to take advantage of others for their own benefit, without exhibiting any empathy for them.


To measure emotional empathy, we used the high_empathy (i.e., compassion) and low_empathy (i.e., callousness) measures from Receptiviti’s LIWC Extension framework. For cognitive empathy, we used the empathetic (i.e., emotional understanding) facet of agreeableness from Receptiviti’s Big-Five framework.


Results revealed that the language SBF used during his interviews came across as significantly less cognitively empathetic than the average CEO’s language during earnings calls. That is, SBF likely does not understand other people’s emotions as well as the typical CEO does. Similarly, results also showed that SBF’s language indicates significantly lower emotional empathy than the average CEO (i.e., SBF’s language expressed lower compassion and greater callousness).



Empathy is an incredibly important factor that helps foster altruism, a construct that SBF has publicly championed throughout various interviews. However, our results suggest that SBF’s language reflects low (cognitive and emotional) empathy in comparison to the average levels of empathy that CEOs tend to exude.


These findings call into question the degree to which SBF genuinely exhibits the altruistic behaviour that he was known for advocating. While high empathy facilitates altruism, low empathy facilitates misconduct. As conveyed in his language, SBF’s low levels of empathy could have served as a warning sign.


Social Aversion


Empathy centers on the ability to feel and understand another person’s emotional experiences. It is often considered the “social glue” that attracts people to each other, and it is integral to developing and maintaining relationships. Because empathy is greatly associated with social engagement, linguistic markers of social processes can help inform the presence or lack of empathy. Specifically, if SBF’s language use indicates lower levels of empathy, his language may also incorporate fewer references to social processes.


We measured SBF’s use of social language with the social (e.g., references to people) measure as well as the affiliation (e.g., references to interpersonal interactions) measure in Receptiviti’s Social Dynamics framework. Results showed that SBF used significantly less language depicting social processes and affiliation than that of the average CEO. In other words, SBF comes across as less focused on (relationships with) other people than the typical CEO, which suggests a sense of social aversion consistent with deficits in empathy.



Self-centered with little concern for others


In addition to evaluating SBF's social language, we also examined his use of self- and other-focused pronouns as a means to understand his lack of empathy. Research suggests that empathy fosters prosocial (helping) behaviours, such as altruism, specifically through one’s perception of greater “self-other overlap."


As past work has employed first-person plural pronouns to assess self-other overlap, we used the liwc.we (e.g., we, us, our) measure from Receptiviti’s LIWC framework to analyze the degree to which SBF exhibited a collective-focus (i.e., perceiving himself in relation to other people). In contrast, we used the liwc.i (e.g., I, me, my) measure, also from Receptiviti’s LIWC framework, to examine his levels of self-focus (i.e., distinguishing himself from other people).


Our results demonstrated that SBF used significantly higher rates of I-words and significantly lower rates of we-words than the average CEO. These findings support the notion that SBF is much more self-focused than other-focused and complement his language patterns that reflect low empathy and social aversion.



Interestingly, other research shows that people in positions of power (like CEOs) tend to use higher rates of we-words, whereas those of lower status and power tend to use higher rates of I-words. SBF exhibits the opposite linguistic profile of a typical CEO.


While studies implicate low levels of empathy in psychopathy, those scoring higher on this dark personality trait also tend to manipulate and deceive others in ways that present the appearance of empathetic concern. Considering that language use is an unconscious behavioural process that can be hard to manipulate, our results demonstrate how linguistic analyses can reveal deceptive behaviours. That is, even though SBF has spoken on the topic of altruism many times, his implicit language use suggests a lack of empathy and a penchant for social disengagement.


Comparing SBF to a crypto industry peer


Although our analyses illustrate naturally occurring patterns in SBF’s language and help provide a glimpse into his personality, there are limitations that should be recognized. For example, our analyses compare SBF’s language during interviews to the average CEO’s language during earnings calls. Interviews and earnings calls are contextually unique sources of language; therefore, differences in language use between SBF and the average CEO may be due to the differing contexts in which language is being spoken.


To help account for this contextual confound, we ran follow-up analyses comparing SBF’s language in ten interviews to that of Coinbase CEO Brian Armstrong’s language in ten interviews. We selected Brian Armstrong’s interviews for comparison because, as a CEO of a cryptocurrency company, he is similar to SBF. However, Armstrong does not possess the same history of criminality as SBF, allowing for a comparison between the two CEOs with less contextual noise that could potentially skew results.


When comparing SBF to Armstrong, results showed the same pattern of findings as the analyses conducted between SBF and the average CEO. Specifically, SBF’s language came across as lower in empathy and social processes (e.g., lower affiliation, lower rates of we) than Brian Armstrong’s language, supporting our initial results.


Regardless, our findings are preliminary and necessitate greater exploration with a larger sample of SBF’s language to further establish any conclusions and generate more nuanced insights (such as longitudinal language patterns that reflect SBF’s mental state pre- versus post-FTX collapse).


Comments


Trusted by industry leaders:

Subscribe to the blog

bottom of page