Data-gathering patterns are far beyond even the most insightful human observers, and the bases of the fortunes of tech giants Facebook and Google, writes Nicholas Agar
In pre-pandemic times a friend running a New Zealand web design company received a job application from an American IT professional. Knowing little about him, they went to Facebook. There they found a 'like' for Donald Trump’s Facebook page. And that was the end of serious consideration of him.
In its early years Facebook made much of its power to connect geographically and culturally distant people. Today’s big stories are about what happens when you add machine learning to those connections. The American job applicant probably didn’t expect to be ruled out of a job in New Zealand because he once clicked 'like' on a Facebook page. But this story hints at how machine learning is increasingly being applied to our digital footprints.
Put simply, machine learning finds patterns in data. In many cases these patterns are far beyond even the most insightful human observers. They are the bases of the fortunes of tech giants Facebook and Google.
In 1993 the cyberpunk author William Gibson observed about technological innovation “The future is already here – it’s just not very evenly distributed.” The same applies to understanding of the economic value of our clicks and likes. The 2021 net worth – USD$119.2 billion – of Facebook’s founder Mark Zuckerberg suggests an excellent understanding. Not so some politicians who should have power over his company.
Responding to a 2018 question from the Republican Senator Orrin Hatch about how Facebook could possibly make money if it were free to join, Zuckerberg complied with advice that, when responding to the questions of politicians who have the power to regulate you, be as boring as possible. His response “Senator, we run ads” is now the stuff of legend. Zuckerberg couldn’t resist a smirk after delivering that line.
What’s the problem if the very clever people at Facebook and Google become billionaires? They’ve given us so many new 'friends' and useful Internet searches. One problem is that the fortunes derive from technologies that many of us are belatedly finding obnoxiously intrusive.
A 2017 book Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us About Who We Really Are by Seth Stephens-Davidowitz, a data scientist formerly at Google, explored the kinds of conclusions that applying machine learning to your clicks enables Facebook and Google to draw about you.
Suppose you like Harley-Davidson motorcycles or the personal care and beauty products of Sephora. You are likely to be less intelligent. Alternatively, suppose you like thunderstorms or curly fries. You’re likely to be more intelligent.
These startling conclusions were based on a study by the University of Cambridge Psychometrics Center and Microsoft Research Cambridge that correlated the Likes of 58,466 Facebook users with scores in IQ tests.
Doubtless there are some Sephora-wearing, Harley-riding geniuses. It would be absurd to say that becoming a Yamaha-riding, consumer of curly fries might be a way to boost your intelligence.
But these kinds of inference about you are the bases of social media fortunes. They are at the heart of the digital tech and social media business model. You may be insulted to learn that your liking for Sephora products supports an inference about your intelligence. But this is the way Facebook and Google are selling you to their advertisers. Their market valuations suggest that they are more often right than not.
Data ethics at Carnegie Mellon University
I’m a New Zealand philosopher who has long focused on the ethical consequences of new technologies. I recently crossed the ditch to Australia to be able to collaborate with Carnegie Mellon in Australia on a course that trains people who work with data on its many moral pitfalls. Carnegie Mellon is ranked by US News Report the number 1 in Information Technology so these are great people to work with.
I’ve suggested that people are unaware of what their clicks may inadvertently show about them. There is a similar blind spot when it comes to the government agencies that we entrust with our sensitive data and data agencies harvest from multiple sources. Organisations including the Inland Revenue Department, the Accident Compensation Corporation, the Ministry of Health, and New Zealand Police have an increasing amount of data about us. They certainly don’t share the profit-making goals of Facebook – IRD doesn’t want to sell you to advertisers. But there is nevertheless increasing awareness about the huge and growing quantities of sensitive data controlled by Government agencies. Data gathering and mining can have a momentum of its own which creates a risk of finding out too late that something we were doing offends both foreseeable and unforeseeable community sensitivities. Even people who want to benefit us can go wrong when data analysis overlooks racial prejudices and, worse still, these prejudices can’t be challenged because they are baked into the data.
Our course offers much-needed instruction to data professionals in Government and business on how the gathering, storage, analysis, and exploitation of data can go wrong even when it is done with the very best intentions.
