When convicted sex offender, alleged sex trafficker and financier to the rich and famous Jeffrey Epstein was arrested and subsequently died in prison, there was a sense that some skeletons were about to come out of the closet.

However, few would have expected that the death of a well-connected, social high-flying predator would call into disrepute one of the world’s most reputable AI research labs. But this is 2019, so anything can happen. And happen it has.

Two weeks ago, New Yorker magazine’s Ronan Farrow reported that Joi Ito, the director of MIT’s prestigious Media Lab, which aims to “focus on the study, invention, and creative use of digital technologies to enhance the ways that people think, express, and communicate ideas, and explore new scientific frontiers,” had accepted $7.5 million in anonymous funding from Epstein, despite knowing MIT had him listed as a “disqualified donor” – presumably because of his previous convictions for sex offences.

Emails obtained by Farrow suggest Ito wrote to Epstein asking for funding to continue to pay staff salaries. Epstein allegedly procured donations from other philanthropists – including Bill Gates – for the Media Lab, but all record of Epstein’s involvement was scrubbed.

Since this has been made public, Ito – who lists one of his areas of expertise as “the ethics and governance of technology” – has resigned. The funding director who worked with Ito at MIT, Peter Cohen, now working at another university, has been placed on administrative leave. Staff at MIT Media Lab have resigned in protest and others are feeling deeply complicit, betrayed and disenchanted at what has transpired.

What happened at MIT’s Media Lab is an important case study in how the public conversation around the ethics of technology needs to expanded to consider more than just the ethical character of systems themselves. We need to know who is building these systems, why they’re doing so, who is benefitting and so on. In short, ethical considerations need to include a supply chain analysis of how the technology came to be created.

The reason why this is important is because technology ethics – AI ethics in particular – is going through what political philosopher Annette Zimmerman calls a “gold rush”. A range of groups (The Ethics Centre included) are producing guides, white papers, codes, principles and frameworks to try to capitalise on the widespread need for rigorous, responsive AI ethics. Some of these parties genuinely want to solve the issues; others just want to be able to charge clients and have retail products ready to go. In either case, the underlying concern is that the kind of ethics that gets paid gets made.

For instance, funding is likely to dictate where the world’s best talent is recruited and what problems they’re asked to solve. Paying people to spend time thinking about these issues, providing the infrastructure for multidisciplinary (or in MIT Media Lab’s case, “anti disciplinary”) groups to collaborate is expensive. Those with money are going to be able to obtain a much louder voice in the public and social debates around AI ethics, and have considerable power to shape the kinds of norms that will subsequently shape the future.

This is not entirely new. Academic research – particularly in the sciences – has always been fraught. It often requires philanthropic support, and it’s easy to rationalise the choice to take this from morally questionable people and groups (and, indeed, the downright contemptible). Vox’s Kelsey Piper summarised the argument neatly: “Who would you rather have $5 million: Jeffrey Epstein, or a scientist who wants to use it for research? Presumably the scientist, right?”

What this argument misses (as Piper points out) is that when it comes to donations of this kind, we want to know how the sausage is made. Just as we don’t want to be drinking coffee that was picked by slave labour, we don’t want to chauffeured around by autonomous vehicles whose AI was paid for by money that helped boost the power and social standing of a predator.

More significantly, it matters that survivors of sexual violence – perhaps even Epstein’s own – might step into vehicles, knowingly or not, whose very existence stemmed from the crimes whose effects they now live with.

Paying attention to these concerns is simply about asking the same questions technology ethicists already ask in a different context. For instance, many already argue that the provenance of a tech product should be made transparent. In Ethical by Design: Principles for Good Technology, we argue that:

The complete history of artefacts and devices, including the identities of all those who have designed, manufactured, serviced and owned the item, should be freely available to any current owner, custodian or user of the device.

It’s a natural extension of this to apply the same requirements to the funding and ownership of tech products. We don’t just need to know who built them, perhaps we also need to know who paid for them to be built, and who is earning capital (financial or social) as a result.

AI and data ethics have recently focused on concerns around the unfair distribution of harms. It’s not enough, many argue, that an algorithm is beneficial 95% of the time, if the 5% who don’t benefit are all (for example) people with disabilities or from another disadvantaged, minority group. We can apply the same principle to the Epstein funding: if the moral costs of having AI funded by a repeated sex offender are borne by survivors of sexual violence, then this is an unacceptable distribution of risks.

MIT Media Lab, like other labs around the world, literally wants to design the future for all of us. It’s not unreasonable to demand that they build it on our terms, not those of a silent, anonymous philanthropist – whether they’re a saint or a villain.

Join the conversation

Does the end justify the means?