Book interview: ‘There Are No Facts’ by Mark Shepard

Way back in the 19th century, Friedrich Nietzsche argued that truth was impossible and the best we could hope for was to work in the realm of interpretation. “There are no facts,” he declared at various times (with variant wordings) in his career, because for the German-Swiss philosopher at least, there was no such thing as ‘absolute truth’.

Here in the 21st century, Mark Shepard has borrowed Nietzsche’s famous words for the title of his analysis of ‘attentive algorithms, extractive data and the quantification of everyday life’. It’s a provocative use of the quotation because we tend to think that our digital ecosystem based on zeros and ones must be ‘true’ to work properly. But you can see why it appealed to Shepard because, that same technology has accelerated the spread of information and data in our post-truth world, where objective facts seem to be secondary to opinion and feelings.

A rare crossover between technology and philosophy, ‘There Are No Facts’ is one of those books that ultimately sets out to make sense of the way we live today. Shepard, who is a professor at State University of New York at Buffalo, says that his book maps the territory of “the uncommon ground that we share in a post-truth world” with reference to machine learning, social media networks and Big Data. He points out that he’s not looking at the sort of machine learning research that goes on in places like MIT’s Lincoln Laboratories, rather “the AI of everyday life. The AI that’s operating on your phone. Siri or Alexa. Or Google Maps developing a cartography of areas of interest which have no relationship to human input or human labelling.” In other words, attentive algorithms that affect the way we live.

One of the reasons for choosing to call his book ‘There Are No Facts’ is to encourage the reader to “focus on judgement”. We live in an age in which it is all too easy to assume that one of the effects of being enveloped by data is that we are also surrounded by truth. And yet, says Shepard, if we look at facial-recognition AI, we are dealing with a process that is fundamentally different to how “human judgement works. With AI you build a set of attributes: gender, age and so on. But we’re also dealing with a world in which nothing is certain,” an observation which Shepard thinks is at the core of what Nietzsche was saying long before digital technology had ever been dreamed of. The problem with object recognition is no matter how much data is fed into the model, “it’s only looking through an aperture that is not necessarily seeing the whole picture”. Human judgement is different, says Shepard, because when we look at people, we base our judgements on attributes outside of that aperture in order to form a more complete picture.

If you analyse the way in which the public receives information, says Shepard referring to former US President Donald Trump’s inauguration attendance figures and the Covid-19 pandemic as examples, “data has become rhetorical”. He reminds me that this is exactly how data started in the Middle Ages when facts were immutable truths handed down by deities. But by the 19th century, scientific discovery meant that data started to have a closer relationship with evidence. What’s interesting about facts, he says, is that they have less in common with their internal validity than with the idea that they are socially assembled. Get enough people to believe something and it will become accepted as fact. Shepard cites authorities that have worked on the contrasting ideas of “on the one hand, the empiricism of the scientific method that produces objective reality, which is unassailable, while on the other there is human judgement that is separate and outside scientific processes”.

This is a concept that Trump grasped instinctively when one of his first acts as President was not to deliver policy, but to manipulate the public’s perception of how well attended his inauguration was (see extract). Let’s be clear, Shepard says, “politicians have lied forever, and they don’t need machine learning to do that. But what we do need as a society is to get better at how we can discern truth claims versus falsehoods.” If fact, Trump’s deception was rudimentary in that all he wanted to do was edit some photographs in order to create the illusion of popularity. It’s just that to get his version of the event to as many people as he could as quickly as possible, he also needed to harness the power of social media. Which is precisely why the echo chamber effect of digital platforms is so potent. As Shepard points out: “Research has shown that tweets containing false statements spread six times faster on Twitter than those containing truthful ones, and contrary to conventional wisdom, it is primarily humans, not bots, who are responsible.”

There’s a correlation between the concepts of post-truth and the digital age, says Shepard. In the past, when we made photographs using chemical solutions reacting with light-sensitive paper, the results had an ‘evidentiary’ quality to them, simply because it was much harder and more expensive to create fakes in the analogue space. Now we can digitally manipulate images on our phones, the expression ‘the camera never lies’ is rendered obsolete. But Shepard is careful to express his belief that the relationship between post-truth and digital technology “is not necessarily a causal one. I don’t hold a technological determinist position that says technology is responsible” for dishonesty. “What I do say is that propagates and accelerates it.”

While Shepard doesn’t challenge the notion that it’s simply much easier to spread misinformation further, wider and quicker over digital networks than it ever was in the pre-digital age, the ‘real problem’ is a much more human one: that matter of judgement that his book concentrates on so much. He’s talking about confirmation bias, “which is all too human”. It’s the tendency we have to assimilate information that conforms with closely held beliefs regardless of its authenticity. While the likes of Twitter “enable the proliferation of misinformation”, the platform itself “remains agnostic”. As Shepard says, the mode of distribution is not the lie, but it does increase its reach, creating “a nightmare in which people don’t know what to believe”.

‘There Are No Facts’ by Mark Shepard, MIT Press, £22.50.

Read our review here.