The Quantified Self: Learning to Observe is out today!

When Kevin Kelly and I organized the first Quantified Self meeting in 2008, self-tracking was very much a niche activity.

Today, self-tracking is everywhere. But we could get so much more out of it!

The Quantified Self: Learning to Observe presents the best of what we’ve learned from 18 years of work helping people bring a scientific mindset to questions in everyday life. I hope you’ll use it to learn more from your own data, and make discoveries that are truly meaningful.

Teaching through examples

We call the application of empirical methods to questions arising in daily life “personal science.” Personal science is best learned by example. The examples in our book follow the format developed in the Quantified Self community: each story is presented directly, in the first person, and explains “What I did, how I did it, and what I learned.” For instance:

• A combat veteran records and investigates unexplained episodes of panic in civilian life.

• A young engineer tracks and uses shifts in wake-up time to help observe recovery from concussion.

• An athlete struggling with bloating and fatigue does a multiphase gluten trial.

• An office worker assesses the fine-grained effects of microdosing.

Below is a scan of the table of contents. On the left is a decorative panel built from our archive of over a thousand QS Show & Tells. On the right are the chapter headings. As you can see, the first three chapters outline the method, and the next eight provide guidance about what to track, how to track it, and what you can learn across a wide range of questions: Pain and recovery from injury; mood, emotion and stress; focus and cognition; sleep and dreaming; digestion and metabolism; heart rate and training; and more.

Each individual Show & Tell is accompanied by data visualizations that illustrate the key methods, and that can be repurposed for your own projects.

But along with presenting the method and teaching with examples, the book also has a higher purpose. The Quantified Self makes the argument that science is for everybody, and proves it by showing how to make real discoveries with self-collected data.

Credits!

Although I wrote the book, credit for the ideas in it belongs equally to my colleagues in the Personal Science Research Group: Thomas Blomseth Christiansen, Martijn de Groot, Steven Jonas, , and .

Thank you for reading our book! And thanks, of course, to the world’s best editor, , for helping start this adventure; to my agent at Calligraph, who was there from the beginning; to the brilliant Paul Sahre, who designed the cover and jacket; and to the team at Workman Publishing for helping bring this work into the world. Above all, thank you to everybody who told their story of “what I did, how I did it, and what I learned” at a QS meeting somewhere in the world. Your curiosity and generosity are the source.

Upcoming events

Right now these are all in New York, the San Francisco Bay Area, Los Angeles, and San Diego, but if you have an idea for bringing The Quantified Self to your city or organizing a QS Show & Tell meeting, just raise your hand (or reply to this email).

Here’s the latest schedule:

San Francisco
Book Talk with Kevin Kelly
Thu Oct 8, 5:30 PM
Book Passage, Ferry Building

New York City
QS Show & Tell with Steven Dean
Tue Oct 13, 6:30–8:30 PM
Civic Hall at Union Square

New York City
Book Talk with Nicholas Thompson
Wed Oct 14, 7:00–8:30 PM
Columbia University

New York City
Fireside Chat with Joshua Ness
Thu Oct 15, 6:00 PM
Civic Hall at Union Square

Palo Alto
eWEAR Seminar: Quantified Self for Health
Tue Oct 20, 3:00–4:00 PM
Stanford University

Los Angeles
Book Talk with June Cohen
Wed Nov 4, 6:30 PM
Village Well Books & Coffee

La Jolla
June Cohen’s The Great Remembering, with Gary Wolf
Thu Nov 5, 7:30 PM
Warwick’s

I hope to see you there.

Finally, if you are working on your own personal science projects, please get in touch. I’d love to know what you’ve learned.

We are now nearly a decade out from a trial run by researchers at Memorial Sloan Kettering that tested the efficacy of supporting cancer patients in self-tracking their own symptoms. The self-trackers lived longer compared to the control group. Eric Topol’s comment on the paper expressed the hope many of us had that the results would open the gates of healthcare support for patients to leverage their own observational and reasoning capacity in collaboration with their clinicians:

Most important, this trial emphasizes the role of engaging patients in tracking their symptoms and of generating their own data. Patients have long been neglected for having a more active role in their care, when in fact, given the appropriate tools, they represent true “blockbuster” potential for improving their outcomes.

But these results did not really hold up over time and across conditions. The most disappointing next beat in the story was the PREEMPT study run by Richard Kravitz at UC Davis, which randomized more than 200 chronic pain patients into a control group and an n-of-1 trial group who were supported in evaluating their treatments. There was no difference between the groups in the primary outcome of “pain interference.”

I’ve read most of the papers I could find comparing self-tracking to normal care, and I won’t go through it all here. But I think any fair reading of the health outcomes research would give the same verdict: Maybe? There’s a genuine lack of clarity despite very serious effort.

In this episode of the Quantified Self Podcast, I discuss this with the Personal Science Research Group. What’s the story of outcomes? We strongly believe, based on our own experience supporting people with complex health challenges, that self-tracking can help clarify a personal situation, reason about treatment options, and come up with new and better ideas about the next thing to do. But shouldn’t the effects show up in outcomes?

Listen to our discussion here.

Upcoming events

San Francisco
Book Talk (with Kevin Kelly)
Thu Oct 8, 5:30 PM
Book Passage, Ferry Building

New York City
QS Show & Tell, co-host Steven Dean
Tue Oct 13, 6:30–8:30 PM
Civic Hall at Union Square

New York City
Book Talk (with Nicholas Thompson)
Wed Oct 14, 7:00–8:30 PM
Columbia University

New York City
Joshua Ness chats with Gary Wolf on What AI Can Actually Tell You About Your Data
Thu Oct 15, 6:00 PM
Civic Hall, 14th Street

Palo Alto
eWEAR Seminar
Tue Oct 20, 3:00–4:00 PM
Stanford University

Los Angeles
Book Talk (with June Cohen)
Wed Nov 4, 6:30 PM
Village Well Books & Coffee

La Jolla
June Cohen’s The Great Remembering (with Gary Wolf)
Thu Nov 5, 7:30 PM
Warwick’s

See all

Just to say: this is a week to really miss Brian Dolan’s MobiHealthNews.1

As everybody in the tech world knows by now, elite smart-ring maker Oura filed to go public this fall, with predicted valuations above $15 billion2. Like all companies aiming to sell shares to the public, Oura Inc.3 submitted an account of its finances to the Securities and Exchange Commission. Here I’m going to use some details revealed in these documents to sketch the company’s vision of the future of wearables.

The Oura ring is not the most important instrument for personal science.4 Nor is it the most popular wearable.5 But it is nonetheless a telling example of how toolmakers think about the epistemic terrain; that is, what we are capable of learning from our new observational capacities and how we’ll go about learning it.

The Oura IPO is a great example to work through because, unlike Apple, Google, Garmin, and Samsung, Oura is 100% a self-tracking company. Its makers were focused on quantified self use cases before the company was even founded. Many of us have met them and talked with them. From the start, they spoke clearly and enthusiastically about how self-tracking can empower us to reason more clearly about our own health. Plus, the ring is amazing. The way it combines signal from different sensors to deliver reliable information about circadian rhythms was truly ahead of its time when it launched, and it has maintained its reputation for accurate measurement. I had a chance to follow along a little bit behind the scenes of the company’s research on ovulatory cycling. They took on this challenge with a seriousness and creativity that impressed me at the time — and still does.

Now they are preparing to go public, with the backing of some of the biggest forces in the investment and healthcare world. (Specifically: Fidelity and Eli Lilly.)

So, what’s their bet? Is this a victory for “citizen science for health?”

What is Oura today?

First, the basics: Oura has approximately five million paying customers. Each of them purchased a smart ring, currently priced from $349 to $499. Since October 2021 these customers have also been offered a monthly subscription to the Oura membership. This subscription allows “access to additional features and functionality.”

The additional features that you gain access to with a membership include daily step count, heart rate, and sleep time. These are among the most useful types of background data for learning about your health, and all are behind the monthly subscription paywall. You can see the table in the footnote for a full list of membership-limited data access, but I think it’s fair to say that Oura has one of the most aggressive members-only policies in the industry today.6 Unsurprisingly, 94% of ring buyers purchase a subscription. And about 85% of the subscribers renew their subscription after a year.

This is a good sign, if you are preparing for an IPO. A common knock against wearables is that people use them until the novelty wears off, and then leave them in a drawer. Oura’s data shows that people keep using it, and they value it enough to pay monthly.

On the other hand, the ring probably set you back $400. Without a membership it is little more than a piece of metal. So, once you buy the ring, you can hardly skimp on membership.

Oura makes four-fifths of its revenue selling rings. Its profit margin on rings is about 47%7 but people don’t buy new rings all the time. Its profit margin on memberships is 89%, and the need for membership renews every month. That’s a better business.

It appears that the recurring app revenue, not ring sales, provides the financial momentum behind the IPO.

But that appearance is misleading.8

An Instrument for Self-Discovery

“A Computer On Your Finger”: Petteri Lahtela and Hannu Kinnunen show their prototype of the Oura ring at QS15, interview by Linda Avey.

When I saw the IPO announcement and the estimate by Bloomberg News that the company’s value could go above $15 billion, my first thought was “what an exit!”

The visionary idea of a wearable ring that could measure activity, temperature, and heart rate was first presented to the Quantified Self community eleven years ago by Petteri Lahtela and Hannu Kinnunen. In 2015, the two Finnish engineers gave a talk called “A Computer On Your Finger” and took questions from Linda Avey and Ernesto Ramirez. At that time, they were just a couple of months away from launching a Kickstarter campaign to prove demand and get them to the first version.

The combination of technical and scientific experience of the founders was impressive. Kinnunen was principal scientist at Polar and led the development of the company’s heart rate and activity algorithms. Lahtela came from mobile-telecom and health IT. Markku Koskela was the physicist on the team, specializing in the collection and analysis of optical signals. All three shared a deep interest in “self-knowledge through numbers.” I did a search on the Quantified Self archives, and noticed that Lahtela registered on the Quantified Self Guide in August 2012, long before Oura was founded.

The talk they gave in 2015 is a classic engineer’s presentation. They focused on the technical challenges and took the value of collecting biometric data mostly for granted. A Quantified Self audience did not need to be convinced that they would like to be able to gather sleep data more easily. Nonetheless, it is interesting to note the use cases that come up during the discussion.

These include:

• Increased awareness of physiological readiness for work or competition.

• Understanding the effects of jet lag.

• Being reminded to eat and sleep at the right times.

• Testing everyday hunches against a reliable record; for instance, the effect of a second glass of wine on sleep, or an afternoon cup of coffee.

• Predicting cognitive trouble.

• Doing collective and collaborative research projects; for instance, students in a class studying the link between sleep and grades.

The overall aim of these engineers was to support individual agency and self-discovery, allowing each of us to make more reasoned decisions in everyday life. Over the past days I’ve spent an inordinate amount of time with Oura’s marketing and financial documents, and I can report that the promise of “self-knowledge through numbers” remains central to Oura’s public identity. As the IPO document puts it: “ŌURA began with a simple but ambitious idea: that we could help people understand their bodies well enough to improve and change the trajectory of their health, not just track it.”9

Having seen countless toolmaker startups come and go, most of them vanishing without a trace, I felt a sense of pride that the Finnish engineers who had shown up at a QS conference with a super ambitious technical idea and the beginnings of a Kickstarter campaign were going to get all the way to a public offering on the US market.

I was especially curious to see how the founders had made out. It was a little disappointing to find that all of them had left the company by 2021, long before the seeds of a strategy that would get them to global scale had borne fruit. The company that is going public is American, not Finnish. In practice, this shift started when Harpreet Rai and David Shuman, two Americans, invested in Oura in 2016, with Rai becoming CEO in 2018. In 2021 Rai was forced out after some corporate in-fighting.10

Strategic continuity from 2016 to today’s IPO plan has come from Shuman, now in the role of Executive Chairman. The board ran a formal search for a new CEO in early 2022 and recruited Tom Hale, previously the president of SurveyMonkey. Hale’s work history suggests he’s gotten enough hard knocks in corporate leadership to know what he’s doing. He’s an operating executive whose expertise is in scaling a company through a big acquisition or IPO. He was the chief product officer of the Austin vacation-rental marketplace HomeAway through its IPO in 2011, and the COO until its sale to Expedia in 2015. He then became president of SurveyMonkey for six years, working on the deal that would have sold it to Zendesk for about $4 billion. Hale started looking for other work as it became clear that the Zendesk shareholders hated the deal. Within days of its collapse he was with Oura. (In the aftermath, SurveyMonkey’s value dropped 60%, while Zendesk, after some deeply confused negotiations, was taken private at a deep discount. Meanwhile Hale joined a company whose investment rounds priced it at about $2.5 billion, and four and a half years later is about to take it public above $15 billion.)

Hale’s most important step after taking the reins was to move into US retail. In April 2023, Oura announced that the rings would be available in 850 Best Buy stores, and that about 50 stores in key markets would have an in-store kiosk where customers could be sized in person. I remember when I saw the Oura kiosk at my local Oakland store; I really had to blink a couple of times to convince myself it was real. Amazon and Target followed, then Costco and Walmart. According to the S-1, retail is now half the hardware business. Ring sales went from 1 million in fiscal 2024 to 3.1 million in the nine months to June 2026.

How does Hale’s success at scaling sales of rings and memberships translate to $15 billion of value? Based on the company’s own numbers, and some standard valuation assumptions, the IPO plan asks us to believe it will sell about 85 million rings over the next ten years and end the decade with about 45 million paying customers.

This seems like a lot. Fitbit, the only real comparator in the dedicated self-tracking market, sold about 100 million devices in its entire history before the Google acquisition, most of them at a fraction of the price and with free access to data. Oura is aiming for more paying members than Fitbit ever had free ones.11

The Epistemic Engine

OK, so what? If Tom Hale and David Shuman and Fidelity and a bunch of limited partners of firms who bought in along the way are going to make money selling shares in Oura to the public in a deal that entails a certain amount of risk — well, isn’t that what markets are for? Speculation supports invention in all kinds of ways, large and small.12 The future is hard to predict. Oura’s execs have global ambitions, and they will end their IPO with a clean corporate setup and a billion dollars of liquidity, and I very much doubt they are out of ideas. Caveat emptor, and best wishes to all.

But that’s not the full story, and the remainder is what interests me the most. If my relatively simple analysis of the financial side of the IPO tells me there’s a gap between the proposed value of the company and the sales data offered to support it, that gap is certainly visible to others. In the run-up to the IPO, the execs will pitch their offering to institutional investors and answer questions. A likely question will be something like: “Accepting, for the moment, that you can sell an unprecedented number of rings and memberships over the next ten years, we see the value of those future sales accounted for in the offering price. Where does future growth come from?”13

And Hale and Shuman have an answer: it is not just cash from future sales of rings and memberships that comprises the asset; it’s what they call the intelligence platform that they’re building with their customers’ data. To quote directly from their prospectus:

“Together our proprietary hardware, software, data, and AI drive a self-reinforcing flywheel that strengthens Oura’s competitive advantage over time. Every day of wear generates richer, longitudinal data and member context, expanding our personalized longitudinal dataset.”

An intelligence platform whose value compounds with scale can be used to support many different kinds of products and features beyond the sale of rings and memberships. The prospectus mentions things like wellness support (for instance “meditation guides”), healthcare integration involving blood pressure and other biometrics, as well as treatment referrals, and population-level discovery services for enterprise and research organizations. (Here you start to see why Eli Lilly came on board this summer.)

So while you buy the ring and pay for the right to see the data it collects, Oura uses your measurements to build both a longitudinal model of your individual physiological responses and aggregates the records of all of its users into evidence for generalizable research, which it can monetize however it likes. Here, Oura enters direct competition with Google and Apple. Welcome to surveillance capitalism; or, and I know it ages me to write this, ALL YOUR BASE ARE BELONG TO US.

I don’t resent that Oura proposes to make money supporting biomedical discovery with aggregate data. There’s a tradeoff, and as individual self-trackers we’ve gained something valuable in return. I appreciate that Oura has succeeded where many other dedicated self-tracking hardware startups have failed, sometimes breaking my heart in the process. The ring works; that was the first triumph. It proved itself to early users in the Quantified Self world, and then crossed the chasm to widespread popular adoption, even while priced substantially higher than competitors. The company grew production and sales to reach millions of users, turned an operating profit, gave a decent exit to early investors, and, if the next months go smoothly, will have a chance to scale another tenfold. Oura brought subtle aspects of biological rhythms to mainstream attention, and developed a research program likely to yield a string of important discoveries.

And yet there remains a difficult passage to negotiate, and the difficulty is evident even in the boastful bureaucratese of the Oura S-1. Even if people like wearing the ring, there is still more work to do, because an intelligence platform, so-called, needs more than sensor measurements to feed it.

A wearable that gives a convenient and well-formatted record of steps, sleep, and heart-rate-derived measures like resting heart rate overnight or heart rate variability, is providing background observations. These function as gentle reminders and support for situational awareness; they also give you the security of knowing that the data is there for you when you want to look deeper, perhaps because you have an urgent question, a symptom or goal, motivating you to think carefully.

But, because these sensor-based measurements are useful mainly in the background, they tend to escape proprietary control. We care about steps, not “Fitbit steps” or “Oura steps.” You can get a decent sleep and step tracker for $100, a third the price of what Oura charges, and no subscription charge.14 The most reliable expert in the QS community on using heart rate variability, Marco Altini, has shown that you can get a good daily measurement with your mobile phone camera as the sensor.

As background observations drift toward commoditization, the more urgent kinds of self-tracking practice remain stubbornly personal. There is no general answer to a question like: “Why did I start breaking out?” Or, “Can I start eating bread again without getting sick?” Or, “Will these exercises improve my tremor?” Or, “Can I avoid scaling back my training, even though I hurt a bit?” Nor is there any definable-in-advance set of things worth tracking. I’ve learned from my own and others’ experience that even a seemingly simple question like “how has my pain level changed over the past weeks” requires tinkering in the context of daily life to get right. Systems of personalized health advice struggle, still, to get anywhere nearly personalized enough.

So wearables companies, even the very best ones, remain generic, their promises of personalization stuck at the level of speculation: “Every day of wear deepens Oura’s understanding of each member, enabling our models to deliver more precise and personalized insights.”15

If only it were as easy to do as it is to say. I’ve noticed, and not only with Oura, that the difficulty in actually delivering on the promise of personalized insights tends to push the promisers into a somewhat aggressive posture. If only people would listen!

Tom Hale, in a November 2025 podcast interview, said this:

One of the reasons healthcare has failed at preventative care is because there’s simply not enough doctors and nurses in the world to educate people about healthcare and the healthy behaviors, and to create the systems of accountability and the motivations and incentives to get them to change. ... The advent of AI, in concert with something like a wearable that you have on your body 24-7, intimate with every change in your physiology, feeding that machine intelligence (and you’ve already got the supercomputer with the screen to display that information to you), which, by the way, has your attention, because you’re staring at it more time than you should — that mechanism is enough to change behaviors at scale.16

Notice what’s missing from this vision: the curiosity and reasoning power of the individual who has already made a choice to buy an Oura ring. That person, the very person whose intense interest in their health has led them to one of the most expensive and sophisticated self-tracking instruments, has been reduced to a mere receiver of information and external influence, the object and target of an administrative mechanism of behavior change.

And it doesn’t even work. At least, I very much doubt it will work. As I read their public statements, Oura envisions AI support living inside their moat, where it will leverage the customer’s data for context and supply actionable advice. This is hard for me to credit. Not just because such a system would encroach on regulated clinical guidance, on one side, and blur into the complex, barely legible terrain of individual inquiry, on the other. These are conventional challenges. The bigger question is how it would work in a world where apps are dissolving. We now have AI support available to us outside the moat. Our robots are increasingly capable of probing and monitoring and copying and reappropriating our records so that we can have more control over them and ask the questions we want.

For a company that has given so much to quantified self practice, the vision Oura is articulating right now has an oddly 20th century industrial-management vibe. But maybe there is a better version here, already implicit in their own system and perhaps already under construction. A wearables company that bridges the gap between the long tail of personal questions and the generic affordances that allow for scaling to tens or hundreds of millions of users may be the one that delivers the best method for finding things out for ourselves. A ring that plays nicely with distributed intelligence, a research program that reveals new use cases and develops new inference workflows, and a membership community that supports learning over compliance: that’s a billion times better than advice.

Show & Tell: Emergency Temps

I don’t flatter myself that the posting day of the QS newsletter is especially important, but you may have a vague sense that this one was late. It was! That’s because I spent the last two weeks in and out of the hospital with a family member with a ruptured appendix. All was fixed in the end, though there were some scary moments. Below is a very brief show & tell I made from one detail of the experience.

What I did

I kept track of the temperature measurements. Body temperature is one of the easiest and most telling measurements for assessing sickness and recovery. In the hospital, it was taken every few hours by a nurse. At home, we measured it whenever we were curious.

There are four reasons why having the record was useful:

1. Fever is a reliable symptom of infection, so even one out-of-range value can provide an alert in advance of other signs.

2. A “nothing wrong” signal is reassuring. When other signs, like pain or nausea, send up warning flags, a steady temp within normal range is a dose of instant calm.

3. Clinical guidance focuses on threshold values, but clinicians with practical experience will look out for trends even if they don’t clearly cross the threshold.

4. Context affects single measurements: a hot blanket, a warm drink, the rise and fall after a dose of Tylenol, a touchless thermometer you paid too much money for—all create uncertainty. A time series builds confidence you’ve got things right.

Here are some reasons that tracking body temperature was surprisingly tricky.

1. In the hospital, the series of body temp records isn’t viewable by patients. You have to ask for each temp, and typically you get only the most recent.

2. Temps are often taken in the middle of the night, when you want to stay quiet, and they aren’t reported aloud unless you ask.

3. Doing non-clinical measurements using cheap and accurate oral thermometers requires a bit of participation from the person who is sick. That’s not always easy.

I wanted us to have our own record, but I didn’t want any distraction.

How I did it

At home I used an inexpensive drugstore thermometer. In the hospital I quietly asked for the temp every time it was taken, even at night, but I often didn’t record the answer if it was normal and the hour was late.

I recorded the temps by softly repeating what I measured (at home) or was told (by the nurse or CNA) into my mobile phone using the app “Just Press Record”. This app is a one-button affordance on my home screen that begins recording immediately and stores the recordings in a folder. I also spoke the time of the measurement aloud. The iPhone mic is sensitive, so I could speak very quietly. I spoke whatever measurement I saw or heard: at home these temps were in Fahrenheit, at the hospital they were Celsius.

The next time I was on my computer, I asked Claude Code to transcribe the recording, convert all temps to Fahrenheit, and append them to a table where I could review them.

That’s it. No graphs, no charts, no apps except the mobile voice recorder. Simple, although far too complex to have done under these circumstances without AI support for transcribing, converting, and arranging.

What I learned

Why was I glad to have this record?

During a brief discharge that ended in a second hospitalization, the trend gave us confidence to skip going to urgent care (which was the guidance of the advice nurse we reached by phone) and go directly to the ER.

After surgery, a slightly elevated value prompted questions to ask when the doctors rounded in the early morning hours. Having a concrete instance to discuss (“what about this high temp from last night?”) changed the tone of the conversation, eliciting a good discussion about what to look for after discharge. We got to hear an experienced general surgeon speaking with us and three residents about how he would use his experience and knowledge to evaluate the need for another CT scan and further intervention. Now, instead of just “watch for elevated temps and feeling worse,” we had a more nuanced view.

I realize in writing this it may be hard to see the value here if you haven’t yourself been trying to make decisions under the stress and uncertainty of emergent illness. What I’m talking about was not, as the dreadful cliché has it, an “aha moment.”

The benefit was very different. The numbers provided a way to grab onto a relevant fact about what was happening, and to use it to ground a consequential decision and anchor a meaningful conversation.

“Should I go to urgent care or the ER now?”

“How will you assess progress after we leave?”

If you’ve been in the hospital yourself, you’ll know that it is not always easy for patients and doctors to understand and aid each other. We speak different languages. We are subject to different stresses and are accountable on different time scales. For families and patients, time seems to stand still. For the doctors and residents, each room is a stop on a tour that can’t go on too long; extra time in one room squeezes the others. And the nurses and CNAs, wow: in the hospital we were in, they had AI-assisted walkie-talkies hanging around their necks with robotic voices interrupting frequently to demand their attention with requests to “reply or escalate.” To explain how all of these clinicians nonetheless brought so much intelligence and compassion to their job requires a different kind of essay.

For now, and as we leave the hospital, I’ll just note: having specific, relevant measurements to talk through helped us think.

QS News Roundup

Just QS people doing QS things. Please write us a line and let us know what you’re working on — or just say hi.

Eric Daza on “Stats-of-1”

Longtime QS contributor and biostatistician Eric Daza has released a preprint of his statistical primer on digital health N-of-1 studies, which will be published in the upcoming textbook Digital Health Technologies for Drug Development. I was very glad to get this open access preprint. Eric’s work formalizes and extends the practical wisdom he’s gained from doing many of his own self-tracking projects. If you like stats or are developing models, you should definitely look at his work. Here’s my quick summary of four key practical tips:

• Consider repeating and alternate conditions rather than leaning too hard on a single before/after.

• Think about onset and carryover. Exclude an initial “ramp-up” interval from comparisons and insert sufficient washout time between conditions.

• Track plausible time-varying influences—day of week, illness, temperature, workload, prior sleep, meals—and preserve the measurement sequence.

• Note that yesterday’s outcome can influence today’s, so treating every observation as independent can exaggerate confidence.

Maggie Delano on tracker threats

Maggie chairs the engineering department at Swarthmore and has done numerous fascinating self-tracking projects. They are lead author of a new IETF document that will shape how the mobile phone will warn when a tracker is following you. Both Apple and Google had separate tracker alerts before joining forces to establish a standard. Now the standard is before the IETF and Maggie’s work on threat models will help guide its development. The narrative scenarios are especially revealing in this document, showing the attack surface these trackers expose.

Martijn Aslander’s Information Autonomy

Martijn Aslander helped organize the very first QS conference in Europe, and continues to be a leader in the free-spirited Dutch community of makers and thinkers about technology. He has a new, highly personal book about regaining agency in digital culture. It is currently available in Dutch and English. See also his reading guide.

KQED on History of the QS Movement

Gabriela Glueck, a radio reporter for San Francisco Bay Area’s flagship public radio station, did a story on the history of the Quantified Self movement, interviewing many of the early participants.

Reprise: Did you miss “Tracking Tics?”

If you haven’t heard my colleague Martijn de Groot talk about his remarkable project tracking one year of involuntary tics, you can listen here.

Fall events

The book launch this fall is giving us a chance to do a series of QS Show & Tell events and author conversations. Most of the events are walk-in, but the two New York events require registration and will likely sell out, so do get yourself registered if you want to come.

San Francisco — Thu Oct 8, 5:30 PM
Book Passage, 1 Ferry Building. In conversation with Kevin Kelly. Free admission, open seating. Event page

New York — Tue Oct 13, 6:30–8:30 PM
QS Show & Tell at Civic Hall, Union Square, hosted by Steve Dean. Tickets $15, or $48 including a copy of the book. Register

New York — Wed Oct 14, doors 6:30 PM, program 7:00–8:30 PM
The Brown Institute for Media Innovation, Pulitzer Hall, Columbia Journalism School, 2950 Broadway. Hosted by the Tow Center for Digital Journalism. In conversation with Nicholas Thompson, followed by Q&A and a signing. Register

Stanford — Tue Oct 20, 3:00–4:00 PM
Stanford eWEAR seminar, “Quantified Self for Health.” Paul G. Allen Building, AllenX 101X, 420 Via Palou. In person and streamed. Register for the stream · Event page

Los Angeles — Wed Nov 4, 6:30 PM
Village Well Books + Coffee, Culver City. “Can We Learn to Live Better?” with June Cohen (The Great Remembering). A ticket includes a copy of the book.

Full list: quantifiedself.com/events

THANK YOU FOR READING
have a great weekend
& drop us a line!

Joining me: Martijn de Groot, Radboud University Medical Center; Sara Riggare, Uppsala University; Thomas Blomseth Christiansen, cofounder of TOTTI Labs; and Jakob Eg Larsen, Technical University of Denmark.

You can see Martijn’s data visualizations here: A single-case (N-of-1) study of Tourette syndrome using personal science to support self-care

A few highlights

Martijn de Groot: “When I recorded them for a couple of months and I looked at the data, I was also shocked by the number of tics that I recorded.”

Jakob Eg Larsen: “Because you have to do this physical thing of tracking, which means you have to have a small moment of awareness which may lead to a small reflection, which may lead to some learnings.”

Thomas Blomseth Christiansen: “You start tracking and you have this steep learning curve. You learn a lot and then it kind of levels off, and then it doesn’t make that much sense to keep on tracking.”

Sara Riggare: “I’ve talked to so many fellow patients with different kinds of degenerative conditions saying that they don’t want to track — I’m just tracking something that I get less well at, I’m tracking my decline.”

For more on personal science and the Quantified Self community, visit the Personal Science Research Group and the QS Show & Tell archive. Martijn’s project, along with many other stories and specific tips on tracking subjectively noticeable phenomena, appears in The Quantified Self: Learning to Observe, which you can pre-order now. Although the official release date is Oct. 6, pre-orders are super-important for convincing booksellers that this topic matters, so if you think you will want to own a copy, please help us spread the message by ordering early.

Over the last two weeks, I’ve had the extreme pleasure of watching a lot of old Quantified Self Show&Tells. Whenever somebody has replied to our pre-order announcement that The Quantified Self: Learning to Observe was going to be published on October 6th and mentioned that they’d been following and participating in the QS community, I searched for any talks they might have given. I found many that I remembered and many I didn’t.

Some would have been impossible to remember because they took place at meetings I didn’t attend. Others were at conferences that I organized or helped organize, but those conferences often had over 50 talks and presentations over the course of two days, so I only got a chance to see a fraction of them. Of course, in writing the book, I watched many videos and read many transcripts, but our database has over a thousand records.

Besides, I like this approach better - I see somebody’s name, and I think “they must have given a talk.” Then I watch and am amazed. Topics that are just now emerging into the mainstream, such as continuous glucose monitoring for non-diabetics and tracking hormone levels and cognitive factors, were already being explored in ingenious detail.

Several important reasoning patterns that are taught in The Quantified Self appeared very early in the records of these talks, and before the end of this post I’m going to talk about one that is seen again and again but rarely easily customizable in self-tracking dashboards. I’m also going to link to a demonstration I’ve made with my colleagues that allows you to explore this pattern using your own self-collected data. If you don’t care much about the background you can go there now and ask your robots for help. What follows is the human bit.

The idea of reasoning patterns is adapted from the work of Christopher Alexander who introduced it in his classic work about proven solutions to challenges in the built world.1 Back when I was just starting out at Wired, this idea was beginning to spread into software development. In the spring of 2000, I got an assignment to go talk with Alexander at his rented house on Shasta Street in Berkeley, about a half hour walk from campus. The exterior was pink and dusty, the interior a desperate mess, and I felt an immediate sympathy for this stout and rumpled genius, who stood and spoke with me in an oversize blue shirt hanging loose and bare feet. There wasn’t a way for the two of us to sit down because the middle of the room where he brought me was taken up by a large model on a high platform, with a hole in the bottom into which he asked me to put my head so my eyes were just above the level of the floor.

Perhaps “talked with me,” is too gentle. He felt I needed firm correction. He’d read something I’d written about Rem Koolhaas and called it immoral. I didn’t like that. I said I’d neither excused nor glamorized. He made a sound of pain. He compared Koolhaas to Adolf Eichmann and asked: “Can one write objectively about such a person?” He had serious doubts about me, and before we agreed to talk more he wanted to understand where I was really coming from. That made sense. But in the next hours, he elicited very little information because he spoke without interruption. The book he was then working on, The Nature of Order, carried his hope — against admitted improbability — that a new way of thinking would transform architecture, along with the civilization that made such brutal and thoughtless design inevitable. At the end of the afternoon, when he tired a bit, I told him how much I liked A Pattern Language and The Timeless Way of Building, and gave him fair warning that the unavoidable question for a man who has built so little and spent so long on an unpublished work meant to overturn civilization will be: is he just a dreamer?

“I’m very vulnerable,” he replied. “To be announced as a dreamer would kill me.” We had a second meeting planned, but both of us understood the risk and let it drop.

The first volume of The Nature of Order came out in 2002, and I treasure my copy. Alexander was part of an emerging resistance to planners and architects who were overconfident about their reasoning power. Coherence, legibility, elegance: these virtues have something going for them. They are scale invariant: a box in a rectangle that is made out of aggregations of other boxes and rectangles, and filled with objects shaped like boxes and rectangles, packs neatly everywhere, including in our heads. We understand the principles in general, and then we apply them to specific cases. But along with this big thing going for them, the methods they spawn have many small things against them. Everywhere they are applied they encounter facts about the world that were not born of high-level abstractions but of extremely low-level processes. The world is full of knotty entanglements you can’t cut through with a knife; though you can cause a lot of harm by trying. For Alexander, modernism was worse than error, it was apocalypse.

The Trainer of LLMs

If you don’t know already about what happened — and it’s hard to predict who knows what these days — the events by which these ideas influenced everybody’s life will seem fantastical. In the 1980s Kent Beck & Ward Cunningham had developed a framework for object-oriented programming based on Alexander’s patterns,2 and in 1994 Cunningham created a website called the Portland Pattern Repository to catalog examples. The plan of the website itself was a design challenge to which Cunningham applied Alexandrian ideas. In March of 1995 he created a supplement with pages that could be edited by anybody, permitting a stable structure to emerge out of a continual process of repair. He called it the WikiWikiWeb. And, yes: that’s the origin of Wikipedia, trainer of LLMs. This is a tragic fate for the turn-of-the-century’s greatest cultural achievement, since LLMs destroy the social process that created them. (Chatbots, as we’ve been taught by hard experience, don’t allow anonymous but secure attribution, deterministic reversion to earlier states, or reward participation; how quickly we unlearn!)

Alexander saw where this was going when he addressed a conference on object-oriented programming in 1996.3 He asked the programmers to stop thinking about his ideas as a mere grab bag of solutions. Yes, there are patterns in the design of buildings and cities. And yes, there are patterns in the design of software. Cataloging these patterns so they are at hand when needed is useful, but not useful enough. What we need are generative patterns, that is, patterns that can be used as part of a living social process to solve problems for ourselves rather than to execute orders. Here’s a quote:

Please forgive me, I’m going to be very direct and blunt for a horrible second. It could be thought that the technical way in which you currently look at programming is almost as if you were willing to be “guns for hire.” In other words, you are the technicians. You know how to make the programs work. “Tell us what to do daddy, and we’ll do it.” That is the worm in the apple.

What I am proposing here is something a little bit different than that. It is a view of programming as the natural genetic infrastructure of a living world which you/we are capable of creating, managing, making available, and which could then have the result that a living structure in our towns, houses, work places, cities, becomes an attainable thing. (my emphasis)

Can you tune your ears right to catch the signal?

Self-experiment and pairwise comparison

In his later work, Alexander emphasized empirical self-assessment of our emotional responses, and insisted that their so-called subjectivity made them more rather than less reliable. In the passage below from The Nature of Order, you can see, somewhat entangled in his typical jargon, a very practical suggestion for how to solve a tricky problem of quantitative measurement of subtle emotional reaction to objects.

Our responses to things, Alexander writes, do not live in the patterns, but in the world of our shifting relations, and therefore there is no way to build a map in advance that securely connects some detail of a design to its expected result. He liked to model at full scale, and, when that was impossible, to use tricks of perception (like when he had me put my head through a hole in his toy floor), to disturb the misleading abstraction that comes from a God’s eye view. Though he was arrogant in argument, his methods were the opposite: simulate, test, reflect, use, and repair.

I’m glad I never wrote anything about Alexander while he was alive. He was too easily hurt to have a spotlight on him; he was renowned within his own circles and ignored where he wasn’t wanted, so there was really nothing for him to gain. But when I had to really cook up something useful; that is, to figure out how to distill and teach methods for self-research from the real life practices of the Quantified Self community, I found him to be just the right help.

Cycles: Reasoning Pattern 9

Here’s a worked example of a reasoning pattern, taken from our upcoming book, followed by a link to an interactive way to test it out.

Cycles

Our expectations are shaped by cycles: circadian, ovulatory, seasonal, and the social rhythms of work, school, and family. When cycles are strongly rhythmic and mostly the same length, they’re easy to spot; you can frame your timeline chart around one or more cycles and your graph will look like a wave. But cycles can be important even when they’re less obvious.

Visual presentations that explore cycles often take the form of a grid with shaded colors. If rows are weeks and columns are days, weekly cycles show up as vertical stripes around certain days. If rows are years and columns are months, seasonal cycles appear as vertical stripes around certain times of year. Cycles that are longer than expected show up as diagonal stripes running down and to the right; shorter cycles run the opposite way.

When to use: When you suspect that your measurements follow a recurring pattern—daily, weekly, monthly, seasonal—and want to visualize and compare cycles.

The method: Create a grid in which rows represent one time unit (weeks, months, years) and columns represent subdivisions (days, weeks, months). Shade each cell based on the measurement value. Vertical stripes reveal cycles matching your column unit. Diagonal stripes reveal cycles that are longer or shorter than expected.

I’ll resist writing more, since you can just try it.

My weekend was leavened by the release of the Anthropic watermark explainer. I’ve been fascinated by AI detection tools for months and have some longer pieces drafting, but the watermark post is too rich an invitation to ignore.

Treat this as a contribution to the almost incoherent noise generated by arguments over AI invasion of human skills. I think surviving creatively in the next year or two depends on being able to handle this noise.

Some years ago I walked down under the freeway entrance to the Bay Bridge, where Survival Research Labs was tormenting the small section of the San Francisco population lucky enough to have heard about the performance with a display of terrifying pyrotechnics. Before the sirens from the fire department scattered the crowd I experienced extreme anxiety that — among other simultaneous catastrophes — their pile of burning pianos would collapse on top of somebody. It was a deeply irresponsible and inexcusable event that I nonetheless remember with admiration and gratitude. I think about that night and say to myself: so many things had not happened yet. Such things still seemed fun.

With much of what we thought were essentially human skills going into the bonfire, the Anthropic watermark scheme appears as a comic policeman blowing his whistle while his pants fall down and Mark Pauline loads a cannon with rotting fish.

Since by vocation I’m a science journalist, I could carefully explain in terms almost anybody could understand why the watermark is more clever and more effective against naive attack than you might think at first, while being so vulnerable to defeat by non-naive attack that they probably shouldn’t have bothered in the first place. But you’ll have to take my boast about being able to do this on faith. Or not; you can equally well disbelieve it. I don’t care, and it doesn’t matter, because the skill at taking a technical invention and making it comprehensible to general readers has been so well mastered by robots that it is hardly a skill at all. I can’t make money from it, and I don’t feel like doing it.[1]

What I do feel like doing is finding a way to express how thrilling it is to watch the most painstaking parts of my professional life become trivially easy. The dissolution of my paranoia about making a mistake now that I have the most powerful forensic apparatus in human history at my disposal has brought me into contact with new problems that —. Well, I was going to say they were previously unseen, but that’s not true. I think the problems that chatbots have exposed were seen or at least felt but they were only approachable as tasks by people with an astounding amount of creative freedom.

A big statement that I won’t try to defend — yet. Instead, let me offer a few scenarios where controlling the ability of robots to alter a text is important in ways that watermarks don’t and can’t address.

Disciplined correction. Chatbots have become indispensable in my work for making relatively sophisticated copy edits, especially (but not only) to references. For instance, what used to be an easily breakable choreography of bibtex imports and Zotero shortcuts now begins with a simple spoken request, transcribed and executed. But, unfortunately, not flawlessly executed. The kinds of errors chatbots make are different than typical human errors, and require a different kind of supervision. As soon as I lose track of what the robots have done, I start to spin. These monsters have to be watched. Often the spans of text are short and the mistakes are not really stylistic. I don’t want robotic contributions dissolved invisibly into my text.

Layering. Any complex analytical work involves layering. For instance, a detailed example that takes hours or days to understand may be compressed into a few lines or even a few words. When I was first starting out as a journalist, I hated losing this material from my drafts. What a waste, I’d think. Eventually, I developed a tolerance for cutting, and finally I loved it. I realized that if I worked through some historical or narrative background well enough I could be confident about the kind of compact formulations — abstract or allusive or metaphorical — that support its conceptual span. Robots layer too, but differently. Their environment is not the world. Their layered constructions need to be probed adversarially. I have to see their work. I can’t have them running loose.

Synthesis and critique. Compose a timeline of key events. List shared features among distinct or competing frameworks. Simulate a particular critical perspective to expose possible flaws in an argument that may otherwise go unnoticed. Chatbots are increasingly useful as synthetic interlocutors. But they are a tricky, translucent mirror. In them, I see my work reflected against the background of their training data. People do synthesis and critique also. Their critique is also tricky. In it, I see my work reflected against the background of their commitments and training and goals and sensitivities. I have to think about the synthesis and critique that comes from humans in a way that’s different than I have to think about synthesis and critique from robots. This is perfectly obvious. On the other hand, how much of the synthesis and critique I get from humans is really coming from robots? if you think watermarks can help with this very deep problem of how we relate to each other creatively, you don’t know very much about humans — or robots.

I can’t say more right now. I want to, though. I think every writer now is being pressed to articulate some facet of their desperate situation.

Note: This text was written using a custom writing surface created by Thomas Blomseth Christiansen that requires humans and robots to take an explicitly named turn before altering the text and preserves the record of each turn. These terms are preserved in a trailer to the document. They are asserted rather than guaranteed, and third parties can have no reliance on them. However, since I am interested in my own relation to robotic influence, I don’t care about third parties for now. You can inspect a version with AI comments marked.

[1] If you want patient instruction on how watermarks work and are defeated, first get an overview from James Padolsey’s interactive explainer, How AI Text Watermarking Works, then point your robots to these references and ask your own questions: John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers, and Tom Goldstein, “A Watermark for Large Language Models.” ICML 2023. arXiv:2301.10226; Sumanth Dathathri, Abigail See, Sumedh Ghaisas, et al., “Scalable watermarking for identifying large language model outputs.” Nature 634 (2024): 818–823. doi:10.1038/s41586-024-08025-4; Nikola Jovanović, Robin Staab, and Martin Vechev, “Watermark Stealing in Large Language Models.” ICML 2024. arXiv:2402.19361; Saifur Rahman Tamim and Amir Labib Khan, “AI Watermark Evidence Fails Forensic Readiness: An Empirical Evaluation.” 2026. arXiv:2607.16010.

Joining me: Sara Riggare, Uppsala University; Steven Jonas; Thomas Blomseth Christiansen, cofounder of TOTTI Labs; and Jakob Eg Larsen, Technical University of Denmark.

A few highlights

Steven Jonas: “By limiting the scope of conveying experience, the power to resonate with people is actually much stronger and wider than if you strove to make your observations apply universally.”

Sara Riggare: “Epistemic agency is the capacity of an individual to act as a producer of knowledge, rather than only a source of data — to pose questions, gather and interpret evidence, reach conclusions, and be recognized by others as a credible knower.”

Jakob Eg Larsen: “We are not talking about people just using these kinds of devices for tracking physical activity, or your step counts. It’s papers where the individual is asking his own question and doing his or her own analysis of the data.”

Thomas Blomseth Christiansen: “Epistemic agency has been an emergent phenomenon. People have just been doing it. It’s not something that was a top-down mandate or an explicit policy.”

For more on personal science and the Quantified Self community, visit the Personal Science Research Group, the QS Show & Tell archive. Also, you can now pre-order the Quantified Self: Learning to Observe. Although the official release date is Oct. 6, pre-orders are super-important for convincing booksellers that this topic matters, so if you think you will want to own a copy, please help us spread the message by ordering early.

The Quantified Self evolved through people sharing their stories. At first with complete informality and spontaneity: I remember at the first meetings we simply put up a whiteboard and had people put down their name if they wanted to show their self-tracking projects.

Later, we started to get into trouble. As the meetings grew, they became a natural place for toolmakers to pitch their apps and devices. This was fun, but it crowded out first person’s stories of discovery that made the meetings so interesting in the first place.

Kevin and I had been part of peer knowledge sharing networks before, and we knew that some gardening was needed. That’s what led to the development of the “QS Show&Tell.” We set aside a full hour before every meeting and put out tables where toolmakers could set up demos, but for the talks, we asked people to always speak in the first person and tell us what they did, how they did it, and what they learned. This turned out to be a great catalyst for years of wonderful talks and presentations.

I want to open the same door here. If you have made your own observations to explore a question in your own life, please consider sharing it. Your discoveries are important.

I learned from watching hundreds of these talks and moderating dozens of meetings that you don’t have to have a big, general, world-shaking conclusion to give something very valuable to the people who read or listen. Seeing people develop ways to answer their own questions with their own data is inspiring, even — or perhaps especially — when the answer remains elusive.

Just tell us what you did, how you did it, and what you learned. And if you have a project that you’d like to share but aren’t sure that it’s ready or how to make it understandable to others, you can call on me for help. I’ve edited many of these, and it’s something I like to do.

Length. 800–1000 words is typical.

Images. Try to include at least one representation of your observations: a chart, a table, or even just a photo of your instrument or notebook. Please don’t bother with illustrations that just decorate, like a big word that says “WOW” or something. It’s hard to really learn something from that. One image is enough, but more is ok, too.

Other people’s data. If anyone else appears in your post please ask their permission first.

Conflicts. If you have a commercial relationship with a tool, company, or product, please disclose it.

How publishing works. Send it to me. I edit. Hopefully fun for both of us.

gary@quantifiedself.com

Hi, Gary Wolf here.

This week, some reflections on the (always simmering) wearables pushback.

Also: a short poll on two subjects, and tuba.

The NY Times on the Dubious Value of Wearables

If you have any reason to add detail to your mental model of how self-tracking with wearables is likely to be understood by a well-intentioned physician, this week’s piece from the NY Times is an excellent primary source: The Silicon Valley Health Trend Making Doctors Nervous: Is more data about your body actually good for you?

Here, Times opinion writer David Wallace-Wells talks to Dr. Rachael Bedard, a primary care physician who has tangled with tech promoters online about the potential value of tracking biometrics without clear clinical value. The standard icons of self-optimization (Peter Attia, Joe Rogan, Bryan Johnson, Casey Means) appear as stand-ins for the practices of self-tracking generally.

We learn about the potential anxiety that comes from aimlessly sorting through self-collected data looking for problems, the risk of false positives and overdiagnosis, and the clueless sense of entitlement that leads techno-solutionists to propose universal benefits from tools that don’t address the fundamental barriers to population-level improvements in health.

Happy as I am to see some of the simultaneously boring and boastful claims of MAHA influencers brought into contact with common-sense critique, it’s nonetheless baffling that this seemingly well-informed conversation completely ignores the challenges of real people wrestling with health questions almost no patient or clinician can address without self-collected data.

Finicky details of dose and symptom management; coping with complex chronic conditions while balancing the demands of daily life; making sound judgments about tradeoffs between health, money, diet, family, and work: self-observation is indispensable for reasoning about these kinds of things. Everybody has questions for which no off-the-shelf answer exists. What are we supposed to do: not reason about them?

The absence of any acknowledgment or evidence of sympathy with the reality of decision making under uncertainty that nearly all people experience at some point was glaring in this discussion.

My Fitbit doesn’t tell me anything I don’t know

Much of the dismissive criticism of self-tracking is directed at popular wearables. And there’s a genuine issue here. A complaint I hear often from people who have dipped a toe into self-research: I have lots of data from my Fitbit/Oura ring/Apple Watch, but I’m not really learning anything I don’t already know.

This is not usually a complaint about data quality. The data is good (enough). The problem is that these devices arrive with the meaning already packaged in. They’re designed to answer a question you may or may not have. If you do already care a lot about what they measure, they work as intended. If you don’t care that much, you wonder: So what?

Uncertainty about the value of step tracking and sleep tracking hasn’t stopped people from buying. (I just looked up the global sales number; it’s hundreds of millions per year.) What’s the point of all these step counts?

I often try to find out if people can get a bit more specific in offering an off-the-cuff explanation of why they track their steps. Mostly, they can’t. Sometimes they say that it’s “motivating.” It seems that many people, including me, my 89-year-old mom, and people I randomly survey in the grocery store line, “just like looking at it.”

I respect this answer. Step tracking is allowed to be fun and interesting for no good reason; maybe it’s like sticking your hand out the car window to feel the rain, or learning the names of plants. Ask me to justify why I do those things, and I’m going to be just as vague. (Although if you press me further, I may try to convince you that humans have what Piotr Wozniak likes to call a learn drive.)

But there’s another reason, aside from basic curiosity, that wearable data is useful even if it doesn’t provide very much insight in its normal app-served form. Wearables give you background data.

The illustration above shows how data collected from a wearable sleep tracker can be used to assess the effects of a change in medication on sleep timing.1 Dots are nightly bedtimes. The line is a two-week average, and the grey band shows the weekly range. Note that while the average bedtime shifts upward by a bit less than an hour, the variation in bedtimes widens. This means: later, but also less regular. The dotted line is the medication change.

I’ve presented this example with an extra strong signal so the method is clear. Your mileage will vary. But questions about sleep and side effects of medication are very common, and having good background data can help you answer them. (Perhaps you’ll see no change at all. Remember: negative results are useful, too.)

The Undramatic Virtues of Background Data

Background data is a concept that emerged from workshops and discussions in the QS community in the years just before the pandemic. At that time, there was a (slight, in retrospect) surge of interest in very small-scale studies and N-of-1 methods among clinical and academic healthcare researchers.2 This interest was welcome, but at the same time there was some confusion about how the practices of disciplinary research mapped onto what people were doing at home in everyday life. After going around in circles quite a bit (including during a phenomenally mixed-up session at an AcademyHealth Annual Research Meeting in Seattle that Martijn de Groot and I tried to lead), we finally got it through our own heads that personal science has some squarely distinct features, and it was only going to make trouble if we kept trying to hammer it into the round hole of industrial-scale healthcare.

One of the most important features of personal science is that it seeks highly specific and immediately useful personal insights rather than general causal explanations. So while biomedicine, as it hunts for causes, speaks of dependent and independent variables, even successful personal science projects may never assert a confident causal claim. Even where some useful insights about cause are obtained, it’s very rare for these insights to reduce to the kind of weighted analysis of variables seen in social science, laboratory experiments, or clinical trials.

No, “dependent and independent variables” won’t do.

And yet, there is something in personal science that rhymes with dependent and independent. When we try to understand something using our own data, not all of the data we gather plays the same role. Some of it represents directly the thing that we care about most. If we care about stomach aches and we track pain and bloating, then our stomach ache incident records are the most important ones to maintain. We want to have a tracking protocol that we know we can sustain for long enough so that the signal comes through. And we want to carefully think about what counts as a stomach ache so we can rely on the quality of our records.

We’ll also be trying to explain our stomach aches. If they started to get worse at a certain time, we’ll ask what else was going on at that time. And our self-collected data can be useful there too, in a different way. Steps, sleep, heart rate, and glucose values from one or more wearables can tell us something about the context or the background of our stomach aches.

I call this data background data because it’s helping to set the frame; and, if something goes wrong with it, or we have some doubts about its relevance, we can often find additional clues elsewhere.

The relation between foreground and background observations is especially interesting when you’re surprised; for instance, if you are unusually tired but you show consistent sleep. Why? The answer might turn out to be important.

I often flinch at the relentless optimization marketing of step trackers and sleep trackers because I think it’s misleading about what they’re good for. No, they won’t cure your sleep or make you fit. But you may really appreciate some background perspective when you want (or urgently need) to think carefully about something in the foreground.

Machine Readable Wishes

Ian Forrester, founder of QS-Manchester back in the day, gave a fascinating talk a couple of weeks ago at EMF Camp on machine-readable wishes: applying the logic of automation rules to the question of what should happen to your personal data after you die. Ian’s talk made me think about my personal archives in a different way. He makes the case that a personal archive should contain executable instructions and provides an open-source framework to help you do this without entangling yourself even further in rent-seeking cloud services.

In case you missed it

The first episode of the Quantified Self podcast went out last week: How to Teach Personal Science, a half-hour conversation with Sara Riggare, Thomas Blomseth Christiansen, and Martijn de Groot about what they have learned putting hundreds of students through the whole question-to-report cycle — sometimes in as little as three days.

Help me make this newsletter better

Any feedback is welcome, but I’m curious this week about two things.

1. Would you like to come to a local meeting? The release of the book may offer a chance to organize some QS Show & Tell meetups in cities around the world. Use the survey to tell me if this interests you, and, if so, where you are in the world.

2. Ravi Karkar, who teaches at UMass Amherst, suggested a channel for academic news/opportunities like conferences, workshops, teaching opportunities, and publications. Should we start a subsection of the Quantified Self Substack for this? (Subsections allow readers to opt out while still subscribing to the main channel.)

   The survey button makes it quick to answer.

QS Reader Questions

PS… tuba

I played enough baritone horn as a teenager to have a lifelong love for the kinds of instruments that make you beg for a ride to school because otherwise your arm will fall off. At the Obama Center opening a few months ago, Damon Bryson, aka Tuba Gooding Jr., showed the world how it’s done. The whole performance is online but I cued it up to the part I’ve played a dozen times. Come for the low brass, stay for the hope.

That’s it for this week.

I hope you like this so far. Reply or leave a comment below.

2

For instance: Sunita Vohra, “N-of-1 Trials to Enhance Patient Outcomes: Identifying Effective Therapies and Reducing Harms, One Patient at a Time,” Journal of Clinical Epidemiology 76 (2016): 6–8, https://doi.org/10.1016/j.jclinepi.2016.03.028; Sunita Vohra et al., “CONSORT Extension for Reporting N-of-1 Trials (CENT) 2015 Statement,” Journal of Clinical Epidemiology 76 (2016): 9–17, https://doi.org/10.1016/j.jclinepi.2015.05.004; Jiang Li et al., “Reporting Quality of N-of-1 Trials Published between 1985 and 2013: A Systematic Review,” Journal of Clinical Epidemiology 76 (2016): 57–64, https://doi.org/10.1016/j.jclinepi.2015.11.016; R. D. Mirza et al., “The History and Development of N-of-1 Trials,” Journal of the Royal Society of Medicine 110, no. 8 (2017): 330–40, https://doi.org/10.1177/0141076817721131; Nicholas J. Schork and Laura H. Goetz, “Single-Subject Studies in Translational Nutrition Research,” Annual Review of Nutrition 37, no. 1 (2017): 395–422, https://doi.org/10.1146/annurev-nutr-071816-064717.

What struck me from talking with people and following links to their work was how effective the Quantified Self Show & Tell format is as a way of sharing knowledge about highly personal empirical practice. People learn best from examples, and although the theory of personal science has advanced, students encountering this form of reasoning about their data for the first time still need to see how it’s done to learn how it’s done.

This is typical of all kinds of science education. We don’t get a sense of research practice until we’ve worked in a lab and seen it firsthand. Something I remember, hopefully not incorrectly, from the chemist and philosopher Michael Polanyi, is that no research program was ever effectively launched without direct, in-person participation from scientists who were already accomplished in the relevant field. Science, he argued, has always been a skill passed on hand to hand.

So, inspired by readers’ encouragement, I decided to spontaneously record a podcast with three of my colleagues who have many years of practice teaching personal science both in a university and in a community setting. I asked them about what they’d learned from working with their students, about the biggest barriers they encountered, about what the students gained, and what surprised them. We even got into some of the details of managing a class that had such high demands on students’ creativity, what kind of homework they gave, and how they graded final projects.

The half-hour episode with our conversation is here. I hope you’ll be patient with a few rough spots in the levels; we literally decided to go for it in a short Slack message and recorded the episode during our regular research meeting the next day.

My guests were Sara Riggare, Thomas Blomseth Christiansen, and Martijn de Groot.

A few highlights

Sara, on what she’s actually giving students and patients: “My role is to give them permission. For the community people, permission to think for themselves; in the university course, permission to think of themselves.”

Martijn, on the “essence” of personal science: “You make structured observations, and the data that represents those observations shows something to yourself that is valuable for you — not necessarily for the rest of the world.”

Thomas, walking through how a vague interest becomes a manageable project: “My area of interest could be my pollen allergies. Within that area, a question could be, ‘What are the dynamics of my sneezes during pollen season?’ And in there, there’s a phenomenon, which is sneezes.”

Nine Tips for Teaching Personal Science

1. It helps to start with a few concrete examples rather than theory. The “what did I do, how did I do it, what did I learn” Show & Tell lets students quickly tune in the flavor of a successful self-tracking project.

2. Explicitly give students permission to pose “small” questions. School trains students to think only big generalizable questions are worth empirical investigation, so they have to get over this barrier.

3. Iteratively workshop each student’s question before tracking starts. Many of them will have to revisit and “de-scope” their question, and they learn a lot from this process. (Though it can be frustrating.)

4. Encourage them to do a short tracking period, even just a few days. This encourages them to do the full cycle, and not get stuck in endless self-tracking.

5. Use freshly collected data rather than existing historical or wearable data. Passive historical data is too “prepackaged” and lets students skip the hard, formative work of defining an area, question, and trackable phenomenon.

6. Cap written reports at a page to a page and a half. A short report is enough to show whether a student has grasped the method, and it makes it possible to give them better feedback.

7. Build in a peer Show & Tell where students explain their project out loud. The act of explaining, not just doing the project, is what makes the experience “become real.” The explaining part is often when they learn the most.

8. Address epistemological anxieties. Personal science requires a different approach than biomedicine, public health, sociology, and other group oriented research fields. For health-trained students in particular, surfacing their randomized-controlled-trial assumptions explicitly helps them stop expecting causal, generalizable conclusions from a single-subject project.

9. Be permissive about the tools students use. There are many ways to turn data into a learning artifact, including pen and paper and disclosed AI assistance. Lowering the technical bar keeps the focus on the personal-science method rather than on data-tooling skill.

Links:

• PSRG (Personal Science Research Group)

• Sara Riggare

• Thomas Blomseth Christiansen / TOTTI Labs

• Martijn de Groot

Hi, Gary Wolf here. That question has been at the center of Quantified Self since Kevin Kelly and I founded it in 2008, after working together during the early years of Wired. We were interested in emerging practices such as life logging, personal genomics, location tracking, and biometrics, all of which added a computational dimension to everyday experience.

While companies were using personal data to understand and influence people, we saw individuals using it to learn about themselves. We wondered whether personal data, like computers before it, could evolve from an institutional tool into something people could use for their own purposes.

The turning point came when Kevin invited anyone who shared our interests to a “Show & Tell” at his studio in Pacifica, California. Thirty people came, many with fascinating projects and extraordinary curiosity. We wanted to do it again, so we created a Meetup group, and soon an informal network of independent local groups sprang up, sharing a common orientation toward empirical self-research.

My book based on these experiences, The Quantified Self: Learning to Observe, is coming out this fall from Workman Publishing. I have five coauthors. All of us have been deeply involved in the Quantified Self community as organizers, researchers, and toolmakers, and this is our attempt to make the practice of personal science accessible to everybody.

That’s a long way around to saying that we’re celebrating the publication of the book by reviving the QS website and archive, and starting this newsletter. Maybe we’ll even manage to organize a reunion conference after publication. Please let us know in the comments or by email if you’d like to be there—and feel free to forward to a friend or colleague who might be interested as well.

You’re getting this email because you are on our Quantified Self newsletter list. In restarting we changed our provider to Substack; I hope you’ll stick with us but it’s easy to unsubscribe if you want.

To begin, here are two books I kept in front of me during the two years it took to bring The Quantified Self: Learning to Observe across the finish line.

Inspiration

These two wonderful books, which differ in purpose, size, era, and personality, have something important in common. Both take a practice thought to require professional training and make it joyfully accessible to any attentive reader.

Mastering the Art of French Cooking has survived the complete transformation of food writing to remain a classic because it never panders to our fears or doubts our capacity to, yes, master techniques that seem intimidating when seen from a distance.

And Ways of Seeing has given a half century of college students their first permission to really look at a painting, and to ask the direct questions that will break their reliance on second-hand opinion: Why is it so big? Are they going to eat that food? Is that a dog’s tail peeking out from under the man’s armor?

My own skills in cooking and looking at paintings took a dramatic leap thanks to these books, and I remember how the weight of my embarrassment in the presence of expertise far beyond me lightened up, allowing me to learn. And then, as I did learn, they became even more interesting, since I was able to pick up on things I hadn’t been able to notice at first.

Empirical research is a skill. Although I consider myself an amateur and a beginner, I’ve been an amateur and a beginner for a long time, and I hope what I’ve written along with my coauthors can help you become a much better beginner, even if you already know a lot.

Books these days are a long shot, because their only purpose is to become a classic. No topic of urgent interest belongs in a book that takes years to come out, and the rising tide of agentic text deflates the value of anything that can be expressed as a list of instructions. A book only matters if it is so coherent that it creates its own world, from the cover to the index.

That’s why I still keep Julia Child and John Berger at hand. I want to feel their influence.

What are the chances my book will be a classic? I get that the probability rounds to zero. But I suppose if you are going to write a book, you should at least aspire to achieve its sole reasonable purpose, and so I have.

For the cover, I had a chance to work with the great graphic designer and illustrator Paul Sahre. Although we were only doing a small project together, with a 6” by 9” size constraint and a tight deadline, Paul’s inventiveness and generosity energized me so much that I still sometimes go back and look at the drafts and mechanicals. Paul wanted to explore as much of the territory as he could, seeing if we could get to that mix of clarity and friction that makes a message stick. I asked his permission to post this video that flashes through some of the things we tried, and of course he said yes.

Paul has designed some of my favorite book covers of all time, including these:

QS Show & Tell

My recent rebuild of the Quantified Self archive of Show & Tell talks gave me an excuse to browse randomly and endlessly. I plan to regularly include one or two that I come across in the newsletter. Here’s one I especially enjoyed because, like the presenter, I was also once a student at Reed College, and had to really figure out for myself how to study.

Separating Work and Home — Lydia Lutsyshyna

Recent Work from QS People

Stats on patient use of AI. Susannah Fox, author of Rebel Health, gives this summary of recent survey research on how people are using AI tools for healthcare. Her many well-informed readers provide additional links. Susannah published some of the first and most important data on consumer use of wearables when she was at the Pew Research Center's Internet & American Life. Her attention here is directly on the most important new thing. General opinion (good/bad, dangerous/safe) is worse than useless right now. We need concrete understanding.

Speaking of AI, Hugo Campos is exploring how agents support agency. He’s created a scoring tool for health AIs based on whether they help people understand, choose, question, and act. (His four criteria for judgment are interesting to compare to the question, observe, reason, explain model we use in our research and teaching.)

I’m going to write more in an upcoming newsletter about using HRV for self-observation, and much of it will be based on Marco Altini’s proven framework. But if you want your dose directly, get his new book about HRV for training: Heart Rate Variability: Science and Strategies for Peak Performance. It is written for performance athletes but contains many general techniques.

You may have read or heard that you should try some supplement to improve your sleep. Maybe it’s a good idea, but how would you know if it worked? Richard Sprague’s Substack, Personal Science, is full of direct, detailed examples of self-research practice. I read it every week. In this recent post he gives an account of his experience with one of the most touted supplements: Does magnesium help me sleep? The post finds an unexpected answer. Even more importantly, it shows you can use background data to reason about immediate personal questions.

Finally, here’s a detailed account from Dana Lewis of her personal heuristics for judging whether certain foods are safe for her to eat. Dana’s post has lessons both for people managing celiac disease and for anybody who has to manage health choices involving endless small details. What do you do when standard published rules and guidelines simply don’t provide trustworthy guidance at the needed level of detail?

That’s it for this week. I’d love to hear what you’re working on. Are you running a personal experiment, developing a new tool, or revisiting an old QS project? Reply to this email or leave a comment below. I’ll highlight more work from the QS community in future issues.