Quantified Self. The search for meaning in our health data

Twenty years ago, Gary Isaac Wolf co-founded the Quantified Self, a movement built around a simple idea: self-knowledge through numbers. Today, smartphones and smartwatches have turned self-tracking into a global culture of measurement and self-optimization, giving us access to health data even doctors may not have. In his upcoming book, The Quantified Self: Learning to Observe, Wolf asks what we can learn when technology becomes a new way to see ourselves.

Quantified Self was established in the same year Apple launched the first iPhone. Did you predict that smartphones would acquire such powerful capabilities to track our health and well-being?

The unforgiving dominance that Apple's concept of the smartphone achieved over computation in daily life was absolutely not predicted by me. By the way, by "concept of the smartphone" I mean that, while we have different brands of phones, Apple's vision of the always-on computer you hold a few inches in front of your face at all times won the day.

My focus at that time wasn't so much on the form factor or the business model of very small computers, but on the capacity they offered for sensing and tracking things that we ourselves defined as important. I very intentionally directed my interest at what I thought of as an alternate way to adapt computing to our own purposes. I was disturbed by the megalomaniac style of social media, with its emphasis on blast radius, and I was curious whether more positive and personal uses were possible.

The Quantified Self movement was once niche; today, self-tracking has become mainstream. But does it still follow the original vision that QS had at the beginning?

The answer to that question is "yes and no."

The diversity and ingenuity of self-tracking practice has continued to inspire me. I think the scientific mindset has spread further than anyone would have thought possible, and many people are really taking advantage of the opportunity to think empirically about their own health. So that's great. I'm less thrilled by the persistence of an outdated way of thinking about computation and population health that sees these new tracking and sensing tools mainly as a way to control and influence people.

I think we should have learned by now that if people don't follow generally well-meant advice, it is not because they haven't been effectively "influenced" but because the advice is not as genuinely helpful to them, in the real context of their life, as the well-meaning advisors think it will be. This mismatch between the administrative goals of the healthcare industry and the health goals of people who are trying to deal with real-life health questions is not even close to being solved.

New wearables equipped with AI can measure an increasing number of biomarkers. Do you think individuals should have access to all this data, or should doctors still manage some of it to avoid misinterpretation? This is often the case with genetic risks.

Misinterpretation is a serious problem, but it occurs within the clinical context as well as outside it. Finding answers that we will rightly have confidence in requires something different than "for doctors' eyes only." But as I write this, I think: the notion of "doctors' eyes only" may itself be behind the times, as it will increasingly be AI-supported systems that attempt to control interpretation. Also dangerous!

Many argue that data and knowledge do not automatically lead to behavior change. Do you agree?

I dislike the term "behavior change" intensely. I see it as a remnant of an outdated model of human agency that nonetheless persists zombie-like in our thinking about health. I prefer "learning."

And with that vocabulary shift, you are completely right. Data and knowledge do not automatically lead to learning. I've seen many instances of people who collect tons of data using wearables and then eventually look back and say, "Wait, I seem to have learned nothing from all this. Maybe the whole exercise is a fraud!" but the problem is that they never developed a personally meaningful question. That's always the first step. If you don't have a question, what good is data?

What is your new book, The Quantified Self: Learning to Observe, about? The title suggests a shift from quantifying to observing. Why did you make that distinction?

Thank you for asking that. Quantifying, whether you think of it as "measuring and recording" or "gathering data," is just not very helpful on its own. By talking more generally about observation, I hope to keep self-tracking in contact with the long history of scientific practice and the scientific mindset, which always involves something deeper than just recording a bunch of measurements and piling up data.

It involves thinking about what particular aspects of the world are worth paying special attention to so that you can gain insight into something you are deeply (perhaps necessarily) curious about. How do you pick what you want to track? How do you design your tracking protocol in a way that makes it feasible in the context of daily life? What adjustments do you have to make as you learn more? Observation involves all of this.

You have argued that personal science is different from conventional science because it is often interested in highly specific and immediately useful insights rather than universal causal explanations. How can we align medical evidence with personalized insights without mistaking opinions for facts?

I don't think there is a hard border between opinions and facts. Instead, there is a gradient of "justified confidence" that we all have to manage. The most important thing is to understand that you need different kinds of evidence for different kinds of decisions. If you are creating treatments for other people or even advocating for those treatments, you'd better have evidence that gives you good reasons to believe your advice won't lead to more harm than good.

If you are attempting to assess how what you do affects yourself, then you are not very worried about generalizability. Instead, you are trying to address your own doubts and understand better how much confidence you should have in your own ideas or intuitions, as well as develop better ideas. Your observations will typically help you here, and it is not very risky. Medical evidence, of course, is very useful at a personal level too, and often it's the effects or side effects of treatment that people care about and are trying to assess. If you are lucky, you have a clinician with an experimental and reflective mindset who can help manage risks and suggest new things.

Before LLMs, users of health trackers got data but had to interpret it on their own. Now AI can combine this data with electronic health records and recent scientific research to offer personalized guidance. Do you think ChatGPT and similar systems will finally solve the interpretation problem that Quantified Self has struggled with?

I think it goes too far to say that LLMs will solve the "interpretation problem." This suggests you can just ask an LLM what something means or what you should do. But LLMs do not have a complete model of your own context and goals, and their advice simply doesn't reach that far.

However, they can do some remarkable things to help. They are almost boring to talk about, and perhaps that's good. For instance, you can give an LLM a printed copy of your medical records and have it construct a comprehensible timeline that you can examine much more easily. You can have your LLMs help you download your wearables data from finicky databases and then ask them to pull out a single data type, like "wake up time," and put it on a timeline chart, with the weekends (or whatever) marked. All of this really tedious arrangement of data can now be handled so much more easily, and that sets you up to think more clearly.

You have recently written about "epistemic agency", the idea that individuals should be producers of knowledge rather than merely sources of data. What do you mean by that?

There is an emerging literature on epistemic injustice that has really helped me a lot in understanding why personal science and quantified self-practice have a special and important role to play in healthcare.

The key point is that people have a need and a right to reason about their own personal situation, and that administrative and bureaucratic systems, including healthcare systems, often make it difficult for people to do this. Sometimes this happens with shocking directness, as when women's pain is dismissed due to blatant sexism; sometimes it happens more subtly, through bias against self-report. And sometimes it is an internal sense that we have that we can't trust ourselves. My thought is that addressing epistemic injustice doesn't simply mean reversing the valence and saying, "Give automatic trust to every self-report." Instead, it means helping, where possible, with tools and techniques that strengthen our ability to reason about ourselves.

In 2027, the Quantified Self movement will celebrate its 20th anniversary. Is self-tracking making us happier and healthier?

That's a general claim I cannot make! But I can say that when people want to think carefully about a question that means a lot to them, there is help available!

Will the quantified-self trend one day become an integral part of longevity, a broader concept that is gaining popularity?

There is, as I'm sure your readers know, a pretty big longevity movement and longevity industry. Self-tracking has played a big role there, but, as always, it's something that cuts both ways. For people who simply are marketing longevity interventions, the rise of self-tracking presents a risk that the evidence people gather in daily life will not back up their claims. On the other hand, perhaps sometimes it will!

Don't you worry that self-tracking mainly makes healthier those who are already health-conscious, leaving behind people without the right equipment or skills?

Yes, I think the wellness culture in general has this problem. I hope that the tools and methods being developed in the Quantified Self community can be very broadly useful. Perhaps one day people trained in medicine and public health will themselves do a self-tracking project as part of their education and share a bit more epistemic agency with patients and clients.

With all the developments in AI, genomics, and wearables, in which direction is Quantified Self evolving now?

I think we are at a pretty big moment. AI is helping to remove a lot of the technical barriers to looking at data that used to keep people from being able to do some of the more useful maneuvers in personal science practice. And, of course, wearables are becoming very widespread. Biomedicine and wellness are both turning their attention to supporting agency. In a very gloomy world, this little bit of good news is welcome.