As news about “fake news” reaches a fever pitch, all the noise is drowning out what I think is a much greater threat, something related but more subtle — and far more destructive: infinite personalisation.
Fake news consists of the absurd, sometimes amusing, and often revolting stories that are easy to spot and dismiss. Infinite personalisation comprises the artificial intelligence-driven, big-data based tools that allow algorithms to build a personalised Internet echo chamber customised just for you, designed to make you feel great. Infinite personalisation feeds you the real, the fake, and everything in between, with the simple goal of holding your attention and getting you to come back for more. It is the process by which companies can measure, match, and predict consumers’ individual preferences with amazing accuracy and then tailor offerings to maximise revenue.
Infinite personalisation is attractive to many types of companies for obvious reasons. After all, if they understand precisely the types of things you want to consume, the probability of making a sale (or getting a click) is greater. For digital goods and services, personalisation tends to be very economical — practically cost-free, in fact, even on a massive scale. And it tends to work well. When it comes to suggesting a streaming music playlist, for example, artificial intelligence does a pretty remarkable job. Importantly, if such an algorithm makes a mistake in this context, no real is harm done; the program learns from user feedback and continually seeks to improve. At worst, the algorithm deprives users of the opportunity to stumble randomly upon something very different that they would really like, resulting in less diversity in one’s musical tastes.
Particularly when it comes to information, however, infinite personalisation has a dark side. Not long ago, Americans tended to watch more or less the same nightly news on TV, read the same newspapers, magazines, and bestselling books. Today, each and every one of us has a custom-designed experience based on our past preferences. Our shopping experiences on major online retailers is designed just for us, individually. We have seemingly endless video choices on streaming media sites, presented to us in a meticulously customised way. Even search engines personalise results.
One result is that Americans seem to be losing a certain commonality of experience — even, in some cases, within the same household. Social network feeds are a fantastic way to keep in touch with your friends. There is no doubt these services are very useful. However, not everyone may be aware that postings in a news feed are carefully selected by algorithms in a very proprietary way, mainly to get us to use the service more. This makes good business sense, but it subliminally impacts our thinking. Like streaming music recommendations, these algorithms are very good at filtering out postings that we’d dislike, potentially robbing us of alternative points of view. This is the echo chamber at work.
Thanks to infinite personalisation, we’re losing commonality of experience just as a record-high 77% of Americans polled by Gallup see the nation as fundamentally divided about “the most important values.” Correlation isn’t causality, but common sense certainly points to a connection. Even with the same raw facts available (to those who would search for them), infinite personalisation likely contributes to increasingly slanted and divided views about the issues of the day. Algorithms will find news (valid or otherwise) and suggest friends likely to validate one’s views, shielding us from alternatives and stealthily reinforcing our differences.
The problem is likely to get worse. The technology of infinite personalisation is getting so good that it’s debatable whether we choose our information sources, or the other way around. Clearly, that’s good business for the providers of these algorithms and the companies that use them to advertise and sell. But it dumbs us down as individuals and weakens us as a society.
When it comes to information about current events (and even history), what these algorithms really are designed to provide us with is pure, mainlined confirmation bias — which itself leads to overconfidence in our own beliefs. It’s easy to argue that these algorithms are simply giving us exactly what we want. They certainly do learn what we like; after all, we keep coming back for more. Further, their designers might argue that the lack of human editors actually removes biases and give us something better or purer than was possible before. Don’t buy it. In the very best of cases, infinite personalisation simply replaces one set of biases (reporters’ and editors’) with another one (readers’).
Does this mean that algorithms themselves can be biased? Definitely. AI doesn’t think, it just scrapes knowledge and ideas from data and finds patterns to present back to us. AI is as biased as the data and rules that are used to train it. In addition, one should ponder this critical question: what exactly is the AI trying to optimise? In many cases the answer is simply company profits, not some broad societal objective.
Combining the insult of fake or intentionally slanted news with the injury of infinite personalisation, what we’re talking about is a vast machine designed purely to earn money by confirming biases on a mass scale, regardless of the collateral damage. It is, unfortunately, a very powerful tool for subliminal manipulation of thinking on a grand, societal scale. It provides a way to manufacture “truth” out of fiction at very low cost and very rapid speed, and the results cannot be anything other than harmful.
The scientific method and the advent of artificial intelligence offer us the promise of greater empiricism and a more evidence-based understanding of reality. Perversely, infinite personalisation seems to deliver the exact opposite. Worse yet, we are still in the early days of this information revolution. The ability to use such tools to shape our thinking will only grow more powerful in the coming years. As individuals and as a society, we should be very wary about their potential to warp our perspectives.
David Siegel is cofounder and co-chairman of Two Sigma Investments, a systematic investment manager with more than $US40 billion in assets.