In the early 1990’s when I was a researcher at AT&T Bell Laboratories, a visitor came to present IBM’s new research on statistical Machine Translation (MT). Those were the early days of statistical approaches to Natural Language Processing (NLP), and several of us were skeptical: translation is hard, and had long been considered one of the holy grails of Artificial Intelligence (AI). History has shown our skepticism to be largely misplaced: while we could not anticipate it then, statistical MT would go on to revolutionize the field, bringing MT back from a pariah status to being a central topic in NLP, leading ultimately to today’s online translation systems that are available to all.
What we also could not know then was that several on the IBM MT team would soon leave to join Renaissance Technologies, applying their statistical and big data skills to the problem of outwitting stock traders, and going on to build one of the most successful hedge funds ever. Much less could we imagine that a colleague of our visitor, who already at that time had a rather idiosyncratic set of political ideas would, a quarter of a century later, use millions of dollars of his immense wealth to threaten the stability of the world.
While the press has been understandably preoccupied with the possible role of Russia in the US election, there has been mounting evidence that Robert Mercer, our IBM visitor’s colleague and current co-CEO of Renaissance, along with his daughter Rebekah, have been instrumental not only in Trump’s ascendancy but also in Brexit. Father and daughter have long been investing heavily in right-wing organizations, one of these being the Heartland Institute, a think tank well-known for its denial of human-caused global warming. Another was Breitbart News, where Robert Mercer was a major backer. Indeed it was through Breitbart that Mercer introduced Stephen Bannon to Trump. Mercer also donated millions of dollars to Trump’s campaign, but his support went well beyond the financial: He helped with logistical matters, engaging the audience targeting firm Cambridge Analytica (of which Mercer was a major investor) in getting Trump’s disorganized campaign on track to win the election. In an independent news story available on their website, Cambridge gloats about their role in the campaign. Mercer had also engaged Cambridge in a similar way for the pro-Brexit campaign.
Mercer’s precise political agenda is unclear in large part because he hardly ever grants interviews. But people who knew him at IBM have told me that his political views even then were quite extreme. What seems clear is that he favors minimal government. Renaissance executive David Magerman (whom I knew briefly when he was an intern at AT&T) was suspended without pay in February for publicly condemning Mercer’s political actions. As Magerman told the Wall Street Journal, Mercer “is using the money I helped him make to implement his worldview [that] government should be shrunk down to the size of a pinhead,” claiming that Mercer “show[s] contempt for the social safety net that he doesn't need, but many Americans do.” Mayer’s recent article in The New Yorker notes that Mercer’s beliefs mirror Ayn Rand’s Objectivism, and again quotes Magerman, who states that Mercer “believes that human beings have no inherent value other than how much money they make,” and that “[h]e thinks society is upside down—that government helps the weak people get strong, and makes the strong people weak by taking their money away, through taxes.”
(Given that view, the connection between Trump and Mercer and Trump and the Christian right wing is striking: “And he lifted up his eyes on his disciples, and said, Blessed be ye poor: for yours is the kingdom of God” (Luke 6:20). But then again, it is no news that the Christian Right has as little to do with Christianity as the present Chinese Communist government has to do with Communism.)
According to the 2010 Supreme Court decision Citizens United vs. FEC, large political expenditures such as Mercer’s are protected by the First Amendment’s Freedom of Speech Clause. While many have decried that decision, as long as it stands there is nothing illegal about any of this. But just because something is legal does not make it ethical. And insofar as Mercer’s support involved not only money, but also technical help from a company in which he was a major stakeholder, one should ask if this behavior transgressed any sort of boundary in professional ethics. The kind of analysis that Cambridge Analytica does involves sophisticated computation on large amounts of data to produce psychological profiles, in order to target one’s campaign to particular segments of the population. In part, then, it uses similar kinds of data analytic and AI methods to those used in some areas of NLP, such as sentiment analysis or opinion mining.
One rare recent public appearance by Mercer was when he accepted the Association for Computational Linguistics (ACL)’s 2014 Lifetime Achievement Award. The ACL is the main international professional organization for people who work in NLP — the technology that underlies MT, and contributes to the spoken language understanding systems that anyone with a smartphone can use. Many of us in the ACL were puzzled by the Award. Certainly it was in part for Mercer’s contributions to statistical MT, but he was just one of several involved in the IBM effort, and he left the field decades ago for Wall Street. But the decision was made, and the ACL is now stuck with it. Embarrassingly, when Mercer is discussed, one often finds a link to a photo or a video of him standing at a lectern emblazoned with the name “ACL”. Not a few of us wish we could distance ourselves from him: this distancing might take various forms, but could be as simple as the Association issuing a statement that indicates disapproval of certain applications of AI-related technology.
Many professional societies have a Code of Ethics. The Association for Computing Machinery (ACM) has one that contains an explicit statement about contributing to society and human well-being. Unfortunately, as of the time of writing, the ACL does not. It is of course an open question whether Mercer’s political actions, to the extent that they have indirectly made use of AI methods, contravened ethical guidelines: that would need to be determined by a committee, who might after all decide they did not, or in any case decline to do anything about it. But since the ACL has no such Code, one cannot even begin that debate.
One moral, then, is a piece of advice to professional societies: if you don’t have a Code of Ethics for your society, get one, lest you find yourself in an embarrassing situation.
But there is of course a far deeper issue here, particularly for the working people who have suffered by an increasingly unjust imbalance of wealth, and who believed that Trump was their voice. It hardly needs pointing out that Mercer is one of the biggest beneficiaries of that wealth imbalance. The evidence seems clear that he played a significant role in Trump’s ascendancy, as well as Brexit, helping both campaigns use sophisticated computational tools to gauge what would be likely to go over well with particular groups of voters — all to engineer an outcome that fits with his own idiosyncratic world view. The Cambridge Analytica page linked above attributes their success to clever psychographics, which in turn uses “a database that the company has created by persuading an astonishingly large number of people to complete surveys on platforms such as Facebook”. Now, more recently Cambridge has claimed that they did not actually use psychographics in the Trump campaign; and a number of experts have questioned the actual impact of Cambridge’s involvement. But the previously-cited page that discusses the Trump campaign and psychographics continues to live on the Cambridge site. And as for the usefulness of Cambridge’s algorithms, while this can be debated, it is a reasonable guess that Mercer would not have recommended them if he did not believe that they would work. Reduced to data points, ordinary working people did indeed figure in these campaigns — but only as a means to an end. They have been sold a very sophisticated, and very nasty, high tech con job.
Again, it is too bad that the ACL does not have a Code of Ethics which it could use to distance itself from the self-centered and self-important Lifetime Achievement Award Winner who helped engineer all of this.
Richard Sproat is a computational linguist and a Fellow of the Association for Computational Linguistics.