As every professional will tell you, one of the occupational hazards of attending social engagements is that people you barely know use it as an opportunity to finagle free advice from you for the price of a drink they don’t even pay for. Doctors get ambushed by strangers looking for second opinions, stockbrokers are badgered for stock tips, and even chefs have to fend off requests for recipes—as if culinary secrets described in a noisy bar can encapsulate a lifetime’s worth of experience.

When people find out I’m a technology lawyer, the conversation inevitably meanders towards privacy. In most instances, there is a 50:50 chance that someone will ask me if it is true that our phones are always on, listening to everything we say. Before I get a chance to respond, someone else chimes in with an example of how they got an ad related to information so personal that it had to be from someone or something listening in on private conversations.

Ad

And things go downhill from there.

A Spy In Your Pocket

According to a Consumer Reports poll, 43% of respondents believed their phone was recording their conversations without their permission. The idea that our mobile devices spy on us all the time is now so ingrained in our minds that it is widely believed to be an irrefutable fact. Everyone has a story about how scant moments after deciding on a family vacation, they’ve been inundated with information on their chosen destination. Or how after returning from a holiday, they got advertisements for a toothpaste they used there, even though all they did was put it in their mouth—not speak about it.

As compelling as the evidence for it might seem, implementing an always-on electronic surveillance system like this is incredibly difficult to pull off. In an article in Wired magazine, former ad-tech entrepreneur Antonio Garcia Martinez crunched the numbers to demonstrate just what a data challenge this is.

The average voice-over-internet-protocol (VoIP) call consumes around 3 kilobytes of data per second—which means that if our phones record everything they hear and upload it to some cloud for processing, each phone will need to push nearly 260MB of data per day to remote servers. Apart from the fact that if our phones really consumed this level of data bandwidth it’s hard to imagine none of us would notice, in aggregate terms, the total volume of data this sort of an operation would generate is too large even for the biggest organizations to analyse. India alone has over 500 million active social media users, suggesting that the ad-tech industry would have to process over 64 petabytes a day in order to target advertisements at us using this information.

Ad

Even if this were technically possible, we need to ask ourselves whether such an approach is a useful allocation of resources. To be able to extract useful signals from the cacophony of casual conversation, we need voice recognition and natural language processing capabilities far in excess of what’s currently available. Human conversation is incredibly complex—filled, as it is, with innuendo, short-hand and nuance—for any existing artificial-intelligence tool to understand. Even if we could record every conversation that takes place within hearing range of the nearest mobile device, we do not have the technology to be able to extract value from it—at least not at a price that advertisers would be willing to pay.

Equally Effective Alternatives

The unfortunate reality is that the advertising industry has other equally effective ways of generating personal insights about us that cost a fraction of what this elaborate eavesdropping exercise will.

Martinez points out that advertisers already pay data aggregators to accumulate vast amounts of data on us from just about everywhere. All this data is peppered with unique identifiers—mobile numbers, email addresses, birth dates—allowing them to cross-reference information from disparate data sources and generate incredibly granular profiles of us. A lot of this data is also tagged with our GPS coordinates, generated by apps we’ve permitted to track our location for a variety of purposes—allowing our personal profiles to be cross-referenced with our family members and co-workers, for instance, in whose physical proximity we often find ourselves. This is how advertisers generate their remarkably insightful predictions about us.

Behavioural Science Explanation

Notwithstanding everything I have said so far, there are probably still some incidents in your personal experience that defy explanation—ad messages so uncannily prescient that there is no way they could have been directed at you without someone listening in. In these cases, all I would suggest is don’t rule out availability bias. As Kahneman and Tversky pointed out, humans have a tendency to weight their judgment more heavily in favour of information more easily recalled. This, coupled with a confirmation bias that is constantly nourished by similar experiences shared by those around, could explain why the conspiracy theory of persistent electronic surveillance is still alive.

The fact is that so long as internet businesses continue to rely on advertising for revenue, data profiling will not stop. In previous articles in this column, I’ve pointed out that we need to shrug off our dependence on advertising and adopt alternative business models:

But there are alternatives – many internet based services already operate on a subscription model. Streaming music services offer vast catalogs of music, podcasts, and a host of other audio formats in exchange for a consumption-agnostic, flat fee just as streaming video services allow us to choose from a wide array of content in exchange for an affordable monthly charge. By taking a small annuity fee from a vast number of users, streaming services can eschew advertising entirely.

Until we do that, I can see no let-up in the inventiveness that advertisers will deploy to understand us more accurately.

Or in the persistence of surveillance.

Epilogue

A couple of weeks ago, in an attempt to improve my fitness, I got myself a new road bike. I’ve always enjoyed cycling but have only ever owned a hybrid. Given the growing enthusiasm for biking in my city (fuelled largely by the pandemic), I was keen to see for myself what the fuss was all about.

As a privacy lawyer well aware of the harms of careless disclosures, I tend to limit the sharing of personally identifiable information to the bare minimum. As a result, I had been careful to make sure not to post details of my new purchase on social media or even mention it to friends by direct message. And yet, barely a week after I made the purchase, I noticed a perceptible uptick in the number of cycling videos I was seeing in my feed.

If I hadn’t been aware of how the online ad industry works I’d have been convinced that my phone was spying on me. What else could have explained the subtle shift in viewing material so soon after the purchase of my new bike? Instead of jumping to that conclusion, I thought more deeply about all I had done differently since I got the bike. And then I remembered that I’d signed up to a couple of fitness apps in order to record the details of my rides. Even though I’d turned my privacy settings way down low, it was obvious that one or the other of them was leaking enough information to result in a change in my internet advertising profile.

Rahul Matthan is a partner at Trilegal and the head of the TMT practice group of the firm. He is a published author and a regular speaker across the world on matter relating to emerging technologies and the law. He writes a weekly column on the interface of law and technology entitled Ex Machina in mint, a leading national business daily.