Is your phone listening to you? | Award-winning LSE iQ podcast

Café noise

CHARLOTTE KELLOWAY:

I recently met up a with a friend for a coffee and while chatting she mentioned a buzzy new bistro in town she wanted to check out so we made plans to visit it the following week for dinner. Later, at home, I noticed an ad on my phone for the restaurant we had discussed. How on earth did my phone know I was interested in visiting this eatery; had it been eavesdropping on my conversation? Welcome to LSE iQ, the podcast where research meets real life. We ask social scientists and other experts one intelligent question – and talk to people directly affected by the issues we explore. I'm Charlotte Kelloway and this month we're asking: is your phone listening to you?

I find out:

What your phone is really doing while you’re asleep

How your employer could access your health status, political views and what you search for online at the push of a button

And why borrowing a book from the library has become an act of resistance

We all have an anecdote of a time when we had a in-person conversation and then received an online advert about something we discussed. I asked Professor Edgar Whitley from LSE’s Department of Management and a privacy expert, if there’s any evidence our phones are actively listening to our conversations.

EDGAR WHITLEY:

It depends what you mean by actively listening. So all modern smartphones, smart speakers, et cetera, do listen out for keywords, "Siri," "Okay Google," "Hey Alexa," et cetera, et cetera. So at a simple level, yes, they clearly are listening because they are designed to respond to those wake-up words and then the next things that you say will become the basis for the actions: change channel, play the next song, order some toilet paper, or whatever it might be.

But I think what lots of people think is that the phone's listening and processing everything that you say rather than at least as far as the manufacturers tell us. If you haven't said the relevant wake-up word, they discard everything that you've said, then they start listening again. You still haven't said a wake-up word, so they will discard it.

I think the reason why people often think that your smartphone is really listening to you and then you'll get the targeted adverts is because the coincidence seems to be so clear that you've talked about a restaurant that you were perhaps going to meet with your friends at, and all of a sudden your social media feed has adverts for that restaurant

….The alternative explanation is the social networks know who your friends are and they know that if a close friend of yours, someone who you interact with on a regular basis, perhaps searches the details of the restaurant, looks for the restaurant on the map or whatever, that it's quite possible that the friendship connection means that things that your friends are interested in, you will also be interested in.

And that often explains why the adverts or whatever pop up. So it looks like it's because you said it out loud, but it's more likely to be the case that….

If your friends are looking up details of a particular restaurant, then the chances are you might also be interested in that particular restaurant. So the advert will be probably relevant for you and therefore it'll appear in your social media feed.

CHARLOTTE:

So, this does make sense to me. My phone - or my social networks - know who my friends are and what they like and therefore what I might like. Carissa Veliz is less convinced. She’s an Associate Professor in Philosophy at the Institute for Ethics in AI at the University of Oxford and author of several books on data and privacy including ‘Privacy is Power’ and ‘Prophecy’.

CARISSA VELIZ:

So companies have been confronted with this question over and over again. And companies tend to say that they don't need that information to get at the information that they want. So they claim that they can infer that information from things like what people search for, what their friends search for, and other data. Which if it's true, that's kind of scary enough in a way. But there's also a reason to doubt that.

First, companies have lied before, and tech companies especially are not the most trustworthy companies I've encountered in my time. But moreover, I have known of cases in which people with a high degree of specialization in on privacy have very particular conversations with very particular language and have recorded that those examples.

And I find it hard to believe that the data was inferred, but I don't know. And that's part of the problem. That someone who specializes on privacy cannot know is a problem in itself. Because I should be able to go to a company and say, give me the evidence. I want to see how you work with the data and exactly what data you collect and or an individual and certainly somebody who cares about privacy or who works on privacy should be able to take their phone and have a way of verifying exactly who's looking at their data and what data and when, and that we don't is astonishing.

CHARLOTTE:

Whether our phones are listening or not may be uncertain, even for privacy experts. However, what’s clear is the sheer amount of data companies collect about us. That data can paint an extraordinarily intimate picture of our lives, often revealing more than any overheard conversation ever could. Just think about it, what was the last thing you searched for on your phone? We often show different sides of ourselves to different people, but our phones see everything: our habits, our interests, our fears, our anxieties. In some ways, they know us better than anyone else. Carissa explains more.

CARISSA:

One of the details that caught my attention when I researched the topic was how your phone is collecting data about you throughout the day and it waits until you charge it, usually at night, to send that data because it's if it sent the data throughout the day, your battery would get drained and you would notice. And so when you first wake up, your phone registers exactly when you wake up because most people the first thing that they do is look at the phone.

And it registers things like how well did you sleep? Did you search for something in the middle of the night? Did you search for something like, I don't know, your mortgage payment or that health issue that you've been having that is worrying you? And it also registers who you wake up next to. Is it your spouse or it might be someone else? And from there, much of the most sensitive data that it gets collected about us is location data, data about where you slept.

And that manages to inform people of where you live and where you work and who you work with and who you live with. And do you drive and how fast do you drive? What what kind of music are you listening to? And what might be inferred from that? If you have a smart car, it's even tracking the weight, your weight, because your weight is relevant to how the car brakes and how it accelerates.

But you can also imagine how that information might be helpful to insurance companies or other kinds of companies that might want to know whether you're losing weight or gaining weight. And whatever you search for online, whatever you buy online, who you talk with, everything gets collected. And it's not only that data that is sensitive, it's the inferences that you can make from that data. So for example, how fast you walk and how stable you walk are data points. Your phone has so many sensors that would surprise you. And from that data, you can infer or try to infer things like what yours what's your life expectancy. Data about how you slide your finger across the screen might tell people some things about you. And and part of what is worrisome is that many of these inferences might be wrong. And you don't get told what data gets collected, much less what data gets inferred.

And what is the accuracy of these kind of inferences?

CHARLOTTE:

Can you talk us through on a practical level how exactly our data is collected and used and by who? for example, is it collected and used by the tech companies? Is it sold on to other companies?

CARISSA:

There are reams and reams of data about you. So of course the tech companies collect your data and part of how they use the data is to improve their product. And that can mean many things. It can mean that they're using the data to train their algorithms. It can mean that they're using the data to study your personality traits and target ads to you, or train their algorithms to work better for you.

And you might think, well, maybe that's a good thing, and maybe it has some good elements but it depends on what we mean of by a system working better for you. Does it mean that it saves you more time and is more accurate, or does it mean that it engages you more, that it just makes you lose time on it more, much like social media encourages you to scroll infinitely. And then there's data that is sold.

So for example, a social media company can collect data about what you're interested in, what engages you, who your connections are, the the times when you engage with social media and much else, and then sell it on to data brokers. And data brokers are companies that amass as much data as they can from every internet user around the world, and they sell on the data in basically two formats. They infer things from that data.

And they have both lists of people, and then in some cases they can sell the file of individuals. So for example, if you're asking for a job to a big employer, that employer might buy your file from a data broker and might learn what you search for, your educational background, your connections, how you use the internet, and all kinds of things. Your political tendencies, your sexual preferences, your health status, all kinds of things that they might not have a claim to access….

…In general, it's precisely our vulnerabilities what is most interesting to companies because they can figure out where we hurt and what makes us tick.

CHARLOTTE:

Ok wow. So data brokers are collecting and bundling together our personal data, which can include highly sensitive information to build detailed profiles on us which they sell to others for profit. These profiles can be used to make life-changing decisions about us, and most of us have no idea it's happening.

Edgar highlights another worrying area where our data can be used to target us. Tailored content and information can be used to skew our world view and political choices. If you’ve ever peeped over someone else’s shoulder at their social media feed on your morning commute, you know just how different it can look from your own.

EDGAR:

Certainly the companies will say you would much rather have adverts that were relevant for you rather than general random adverts….

….But exactly the same kinds of algorithms can also be used to influence what news coverage you get, your perspectives on world events, because we see increasingly that young people are not reading newspapers, watching televised news programs. They're getting all of their knowledge of the world from the social media feeds of the apps that they are using. And as a result of that, they are going to get self-reinforcing, skewed kind of things because the same algorithm that says you like these products, it will show you adverts for that, are not suddenly going to show you adverts for something that you've shown no interest in previously.

So if you've had particular political opinions that you've been particularly interested in, you're not going to want to see... A good person would like to have a whole range of views, but the ordinary person is just going to say, "I still want the kind of news that reinforces my worldview," et cetera, et cetera. And so you can get really distorted unrepresentative things. And in extreme situations when you get contentious votes, Brexit, et cetera, having a perhaps not completely representative news feed, can then lead to people legitimately voting how they want to vote, but perhaps not doing it on the basis of a full set of information, but rather an algorithmically-programmed set of information.

CHARLOTTE:

Twenty, thirty years ago there was a more limited pool of places where we could get our news and information from. Many of us sat down to watch the same news programmes, read the same papers and shared a common set of facts. Sure, different media outlets have always had different political leanings but you could generally always access what other people were consuming, if you wanted to. Today, algorithms learn our specific fears, interests, biases and can curate highly personalised information streams, shaping what we see and reinforcing what we already believe.

Highly personalised content was used to target individuals in the Cambridge Analytica Scandal which broke in 2018.

Here, data scientists harvested Facebook data from around 87 million users through a personality quiz app called This Is Your Digital Life. While only 270,000 people completed the survey, the app also collected data from their Facebook friends without their knowledge. By identifying personality traits and linking these to Facebook activity such as ‘likes’, the company could build psychological profiles of users and micro- target them with highly personalised political adverts, including during the 2016 US presidential campaigns of Ted Cruz and Donald Trump. The effectiveness and extent of its targeting in these campaigns however does remain disputed.

EDGAR:

I think what was particularly interesting about Cambridge Analytica was that if I did a quiz and I therefore gave access to them to my social graph, they also had access to the social graphs of all of my friends, unless my friends had kept their accounts private.

So I might say, "Yeah, this is the account that I don't really use for anything serious. So I don't mind giving some of my data in order for the benefits of the psychology test or whatever it might be." But even if I'd appreciated that and that was a deal I was prepared to do, I would have thought that most people didn't realize they were also dropping their friends in it at the same time.

CHARLOTTE:

As Edgar mentioned at the start of this episode, all of our data is connected. Our phones know who our friends are, what they’ve been searching for and what they like. That information can be used to make assumptions about us too.

Carissa highlights how often our personal data is not so personal.

CARISSA:

I think the term personal data is quite misleading because it suggests that personal data is individual, which often it's not. So if you share your genetic data, you're sharing data about your family as well. And not only your close family, but potentially quite distant family. And it also suggests that there is something of a personal preference in it. Like, well, you know, I like strawberry and you like raspberry, and that's a personal choice. But privacy has to do with politics and with our collective lives.

So when you share your privacy, it's not only that you're probably exposing people you love, including your family, your friends, your neighbours, your colleagues, but also you are exposing citizens around you. Because it's a bit like ecology. If you don't recycle, it's not only that you will suffer the consequences of climate change or whatever it might be, but also that other people around you will too.

CHARLOTTE:

In 2017, the Economist published an article headlined ‘The world’s most valuable resource is no longer oil, but data’

Data has become central to our economy as one of the world's most valuable resources, underpinning entire business models and industries.

This is something Professor Nick Couldry from LSE's Department of Media and Communications discusses in his book Data Grab: The New Colonialism of Big Tech and How to Fight Back, co-authored with Mexican/US academic Professor Ulises Mejias.

NICK COULDRY:

Ulises Mejias and I. We've been working on this framework of data colonialism for about 10 years now….

…And one thing we're definitely not doing, because after all, colonialism was very, very complex, unfolding over centuries, we're not trying to make a crude one-to-one comparison between what's happening today with data and tech and everything that happened in the past four centuries with colonialism…

…What we're doing is we're saying there were two key aspects of what's going on with data and tech today, which we can trace back to the two core aspects of colonialism historically. And we see those as in common, and therefore we could argue continuous with the past.

The first key aspect is the appropriation of vast amounts of resources. In the past, four centuries ago, five centuries ago, it was the taking of the Americas, a continent that the explorers didn't even know it existed. But it had vast amount of mineral resources, unimaginable gold, silver, unimaginable wealth generated. And they felt, why not? They could take everything. They were taking everything.

…That idea of complete capture, the only historical parallel to what's happening today in the data sphere is the colonial conquest of five centuries ago in our argument. But it doesn't stop there because it's also about the mentality with which everything was captured. As we know, the conquistadors who captured the Americas had a sense of entitlement….

…And as Empire developed, the British developed a very smart explanation of why this was absolutely the right thing to do. They argued that in God's eyes, the indigenous people were not using the land properly. They were not making enough profit out of the land, and that displeased God. So it was necessary for the English to take over, to take the land and make more out of the land, and please God in the process.

Now what we have is data barons, big tech companies, arguing that they have a perfect right to take our data….

And so we believe that there are real parallels between this mentality that not only can you take everything, but actually, you have a right to, you're entitled to, and the economy depends on you doing so. And so you should just be allowed to get onto it.

CHARLOTTE:

Nick and Ulises argue colonialism which was historically a land grab of natural resources, exploitative labour, and private property – has taken on a new form where big tech companies control and exploit our data for profit.

So we’ve looked at various aspects of data extraction but how does all this work from a legal perspective? This does depend on where you live in the world. In the UK and EU, we have GDPR - General Data Protection Regulation. This came into force in the EU in 2018 and after leaving the EU, the UK enacted UK GDPR which largely mirrors the EU legislation. It is an important component of privacy law and was brought in with the goal of improving people’s control and rights over their personal data, I ask Nick about the difference it has made and if it goes far enough in protecting people’s data.

NICK:

Well, the GDPR was a very, very important step philosophically, if you like, because it was the first major challenge against the dominant American principle…, if you like, the idea of neoliberalism….

that said market's always know best. And the first sentence of the GDPR says, "The collection of personal data raises fundamental questions of human rights." That is a direct challenge to that principle. And it's really important.

The problem was that it relied on people giving their consent to the giving up of data and the use of data, including the use when it changes and it's passed onto someone else. The problems we were talking about. Now consent, on the face of it, sounds a good principle, but as we know, it's very easy to get your consent when you don't really consent. It could be just extremely inconvenient to slow things down and go through all those boxes. Or your employer says, "You have to consent because you have to be part of the system. Otherwise, you're not going to have a job." Well, that's consent, but is it really consent?

CHARLOTTE:

I’m sure we can all think of a time we’ve clicked "accept" without reading the small print. Many terms and conditions are lengthy, complex and difficult to understand. Although technically a breach of GDPR requirements for freely given and informed consent, some websites also use something called "dark patterns" – these are design tricks that nudge people into handing over more data or making choices they didn't intend to make. This includes tactics such as hidden opt-outs, confusing language including the use of double negatives, default options and ‘confirm shaming’ prompts where to opt out you have to click on a box saying something like: ‘No, I don’t want to save money’ in order to proceed.

NICK:

So there's so many ways of extracting consent. And many legal theorists, as a result, have become deeply questioning about whether consent is the right principle to govern something as complex and as opaque, as non-transparent as the data gathering we've been talking about. That said, if properly enforced, these consent rules can be pretty powerful. And they've been used against Facebook to restrict its data gathering in Europe. And then when combined with more recent legislation like the Digital Markets Act and the Digital Services Act, they've started to, the way I put it, would be throw sand in the engine of big tech pretty effectively in the past five years.

And there are real disincentives now in dealing with Europe in certain ways for the US big tech. But it hasn't fundamentally changed their course, which makes me think that a more radical approach that would perhaps say certain types of commercial surveillance actually should be banned.

CHARLOTTE:

While GDPR seems to have made some important steps in the right direction, its focus on consent can be problematic. Carrissa also questions how effectively GDPR is enforced.

CARISSA:

It gets enforced in the most thin way you could possibly imagine. And so there is the law and then there is the spirit of the law, and then there is the actual practice of the law and those can be quite distant from one another…

Now we are facing situations in which you know people have Amazon ring cameras or meta glasses. Does it mean that I'm consenting because I'm not telling them please stop recording? I might not even see the camera. So there are so many ways in which this system is obviously dysfunctional…..

And then all of this is all the more worrisome because, and there's a connection with the law here. Much of European law is based on the idea of autonomy and much of, for example, medical law….

The idea that people have not only the ability, but the rights to lead their own lives, to have power over their own lives, to decide what values they endorse and to live accordingly. And to not mess with that, to not impinge upon other people's autonomy. But this whole machinery of surveillance is built to feed a machinery of prediction, to try to figure out what people are going to do next and try to influence their behaviour.

CHARLOTTE:

Autonomy means having the power to make your own choices and live by your own values. But today's surveillance systems are designed to predict and influence our behaviour, raising concerns about how much control we really have over our own lives.

Many of us now consult technology to help us make important life decisions. As mentioned earlier, we may turn to the internet and AI to ask questions about our health worries or our mortgages but is this making us more and more reliant on these tools?

his new book Predatory AI: How We Can Resist Corporate Capture of the Human Mind Nick talks about data extraction and how using technology like chatbots can create dependence.

NICK:

Well, as we know, chatbots work by predicting the most likely next words in a sentence or in a sequence of sentences. They just predict. They don't understand you…

…Now in general, there can be good uses of chatbots. They're not necessarily problematic. But when people are using them in a way to ask quite personal things... Which is, we know many people are now doing this, then you have a very different situation. Because the chatbot emerged when a large language model, which is a very impersonal thing; it's just vast banks of servers with complex mathematics and engineering. It's not a friend. It doesn't have a friendly face... Is linked to language that we, as human beings, associate with friendliness, with chattiness.

Then it becomes something that we naturally, as human beings, interact with as if it were like us, even though it's absolutely not like us. As a result, we attempted to share questions, very personal questions, share feelings. There was a study recently by a Guardian journalist of, asked…undergraduate students to give all the information, how they used ChatGPT over 18 months. And he found that they started in the normal way, just asking very practical questions. But by the end of the 18 months, they were living virtually their whole life in dialogue with ChatGPT.

Now, then, as economists would say, there's the opportunity cost. What would you have been doing if you hadn't been using ChatGPT? And of course, what you would have been doing was thinking, talking to yourself when it really gets tough, going out for a walk, working out what on earth do I do now? You'd be jigging on the resources of your own brain, maybe talking to friends who would do the same and struggle. "Well, I think you need to do this."

But now people are short-circuiting that and going to a predictive machine, which gives you, as it were, the average of what humans happen to have said in the training model that's feeding the predictive machine. And so there's no limit, if you like, as to how close we can bring these AI machines to us. There really is no limit. People are asking the most intimate questions. And of course, the more they do so, the more dependent they're becoming on those answers. And this is, if you like, personal data extraction on a much, much detailed level.

Obviously, the data we give in terms of asking a question, responding to the answer we get to our question feeds more into the AI model. So it's extracted data. But of course it's having a much bigger function now that it's training the AI to train us to use it more and more. And this is where the economics of this comes in. The only way the AI industries will make money is to recoup the vast amounts of hundreds of billions of investment they've made in training the AI models. And the only way they can do that is if most of us are using AI tools most of the time as intensively as possible, which means they need us to develop these habits of dependence.

They're an economic necessity. And therefore our previous objections about privacy, protecting that private space where I am alongside myself, no one can get in there. That is me when I'm really with myself. We're surrendering that for no obvious advantage.

CHARLOTTE:

So, going to back to our original question. While your phone is likely not listening to your private conversations, the reality may be more unsettling. The technology that surrounds us constantly harvests data about our habits, preferences and routines, enabling algorithms to predict and shape our behaviour with remarkable precision. They don't need to eavesdrop on us because they already know enough and we are living in a world where, as Nick has just said, we are becoming ever more dependent on this technology.

Sometimes, it only takes a small act to help regain some control. As Carissa notes:

CARISSA:

When you're reading a paper book in a library, nobody can touch you. Digital tech and the big tech companies

Cannot get anything from you. And in in a way, it's turned out to be the most rebellious act of defiance in the digital age.

CHARLOTTE:

This episode was written and produced by me, Charlotte Kelloway with script development by Sophie Mallett and Sue Windebank and edited by Oliver Johnson. If you’d like to find out more about the research in this episode, head to the show notes. And if you enjoy iQ, please leave us a review to help other people discover the podcast.

Join us next month, when Sue Windebank asks, ‘How can we stop buying so much stuff??’ [preview clip of audio from next episode]

If you liked this podcast you might like the LSE Events podcast which features talks by some of the most influential figures in the social sciences.

Listen to a recent talk, for example, by Professor Nick Couldry who we heard from earlier, and Baroness Kidron asking ‘Can human solidarity survive social media and what if it can’t?’

For more inspiring content search LSE lectures and events wherever you get your podcasts.