Wednesday, June 1, 2011
Quantified Self Conference 2011
I had a total blast at the Quantified Self Conference, which was held at the Computer History Museum in Mountain View, California. I gave the Sunday Morning Keynote, on the subject of sensors for QS - you can download my slidedeck: The Future of Sensors. It's image rich without much text, but you can still get a feel for the breadth of my talk - it was all overview and little substance :-). Still I got a lot of people telling me that they really enjoyed my presentation and/or were inspired to get out there and experiment (which was my final message) so I think the talk was a huge success. I also really enjoyed several of the sessions that I went to, plus I met several awesome people.
Some overall comments: the conference wasn't very much like the informal QS meetups. The QS meetups are show and tell sessions, where people volunteer to talk and get a few minutes at the front to share their tracking story, be it health, life style, geeky or whatever. The conference, despite being a little bit unconference-y, was fairly formal, which I think is kind of unavoidable at the scale it was at (400 people, I think?). People signed up on a wiki ahead of time to lead hour-long breakout sessions on whatever interested them. My experience with these sessions was hit or miss - sometimes it was awesome, if the organizer was knowledgable and knew how to manage the crowd and the sharing process. But sometimes it was less than awesome, with a monologue from a boring speaker and a noisy room in which it was hard to hear anyway. Several of my friends also complained that it was too business-y for their tastes, with a lot of sessions focussed on products, marketing, buzz-word narratives, etc. I think that criticism is totally fair, and more than likely it's a result of having charged $200-$400 for the tickets: you squeeze out all the hobbyists and get more of the professionals, and that tilts the conversations (and consequently the sessions) more towards business-type stuff. I have also always felt that QS was a little buzz-wordy and commercial, even at the meetups it's pretty frequent to get someone talking about the app they made and are selling, or the product they are developing (and I'm guilty of this myself, of course). So while I do lament the under-representation of the "true stories" kind of thing that happens at the QS meetups, I totally think the conference still served a large aspect of it's user base, the innovators and entrepreneurs who are interested in getting QS-type stuff out into a larger mainstream. I guess for future conferences it might be better to market it with an emphasis like that - either that or change the conference to be more informal show&tell so that it more closely matches people's experiences of the meetups. The point is to more closely align people's expectations with what's going to happen.
My experience was still hugely positive, and I had a great time. I totally think having the long breaks between sessions was key, to give people time to socialize and check out the posters and exhibitors, which in my opinion was one of the big successes of the conference. I had a great time talking to the exhibitors, and I could totally have used even more unstructured time. I also think the conference could have benefitted from shorter breakout sessions. An hour was a little too long. It would have been better to have 30 minute or maybe 45 minute sessions, and have a few more of them, plus a little more unstructured time. Another big win in my opinion was the Ignite-style talks right after lunch. It was a great way to hear about lots of different things quickly, and the after lunch slot was a good time for it - people really just want to sit and digest and not be too active or make any decisions, so having an enforced time filled with quick talks was awesome.
Anyway, some of the sessions that I particularly enjoyed: openyou.org hardware hacking session (by Kyle of openyou). Galvanic skin response and multi-sensor input (by David Andre of BodyMedia). The programmable self (about behavior change), by Trotter. My full conference notes are below.
Some other stuff I have seen about the conference: Bullet Proof Executive: butter sardines yum view from the quantified self conference, Quantified Self Guide, a devices and services guide website that the QS people launched at the conference, Twitter Insights from Day 1 and Day 2, Ethan Zuckerman's Live Blog of QS - prolific!, Eri Gentry's notes on my talk, Delta Self Roundup of Links from the conference, and finally a tagged flickr gallery from the conference. You can also see my little flickr set.
And now, my complete notes:
Quantified self conference 2011
Purpose of the weekend isn't knowledge transfer but rather thinking
In QS the self is not a little add in - it's the main thing
Even advanced users are confused by advanced users
Advanced users teach technologies to do new things
We differ about means but we share the same ends
What did you do
How did you do it
What did you learn
Why does personal science matter
Self experimentation was really powerful - can find out things the experts don't know about
Part 1 - health care stagnation
Scientific American cover story on obesity "I had to double check that is was a 2011 magazine
Part 2 personal science
Part 3 why is personal science succeeding where professional science is failing?
Rituals interfere with progress - randomized placebo controller clinical trial like a rain dance - done to a high standard, reassuring, but may not produce rain!
Qs business models breakout
Ben Rubin of zeo - sleep management, not a "sleep device"
~100 people, 20% female
Hardware is a race to the bottom - and retail margins hurt a lot
So they created a service to go align with it - hasn't worked very well yet.
Almost no one bought a subscription - some people did buy the bundled sensor plus service - they wanted the best thing available
Now they are back to just the $199 product
Advertising? We have a very well targeted demographic, but it's not that large...
One time service - marathon sleep coaching service...
Affiliate - selling other people related services
Selling the data?
Didn't mention razor blade model - disposables
- apparently their headbands wear out and have to be replaced - but they've tried to not have this be the case - the silver fabric wears off.
The psychology of price point
Education and explanation around the product - not a quick fix solution, so price insensitivity was not as great as they thought
Body media has had huge success with personal trainers - the leading type of purchase is via recommendation
Consumer wellness vs medical device vs fitness device, etc.
Health care less resistant in other countries?
Who benefits from improvements in health? The consumer, yes, but also employers, healthcare companies - can they pay for it somehow?
How much have you segmented your "sleep market"? Frustrated sleeper. Performance sleeper.
Temporary model - many people only use it for a few months - although others think that, but then end up using it more long term, even if infrequently. Loaner price probably only makes sense if price is really high.
Syngeristically sell products, cobranding, bundled offers - this can also help grow the space, we have a lot more in common than we are competitive, and qs tracking movement remains small...
Web service : 1 easy to use 2 engaging, interesting, providing value
Not everything can or should be gameified or socialified, e.g. People don't want to share their cc purchase history. Turns out most also not interested in sharing sleep except with a few super close people.
Intrinsic increase in value of time as people collect and add more data - this kind of increasing value is important for a freemium model
Big data is not big brother
Reflections in a digital mirror - book, iPhone app.
Scanadu - building the medical tricorder
Sequencing human lifestyle
Mecuryapp - what if I could get information from my feeling
When I was a kid I liked to keep a journal, now that I'm an adult, I like to keep charts :-)
Gotta get this app!
Smiling predicts longevity!
Three parts of mindfulness - attention, intention, attitude
Openyou.org - hacking hardware
Stay at the protocol level - expose the data, then let app developers do what stye do best
Early entrants had to build vertically integrated offering hardware and analysis software, but that's becoming less true now - other ways to get better analysis.
Openeyes eye tracking hack
Emg device - gym mate or something like that
Allan - continuous monitoring of glucose via "implant" - actually needle sticker, replace once a week $50 (box of 4 for a month for $200
Mailing list at open you.org, irc channel as well.
Audacity needs you
We all have a serious health problem - death
Qs business panel
Ben rubin zeo
Jason Jabobs fitness keeper
Bryan Cusarak green goose
What are you learning about tracking as a habit through your customers?
Zeo intense about how they use it 3-4 nights per week for 3 weeks to 4 months
Then use down, but 70% still using 6 months out!
Fitness keeper - users who integrate with facebook are more likely to keep using it. Driving reengagement is also easy just send an email about setting a goal.
Green goose - can't fall off because the devices are ubiquitous and ambient. Persistent passive sensors.
Green goose product launch in a fall timeline :-(
Zeo has the largest database of sleep stages in the world - by a couple of orders of magnitude!
What's the value of helping millions of people sleep vs doing science on that data? Totally not clear.
Nobody is doing stress - missing!
Jason very positive about future of this stuff - all the building blocks are in place, just need time progression to make it truly mainstream
Quantified self conference day 2
Galvanic skin response and multi sensor input
David Andre of body media
2 kinds of sweat glands in the body - cooling and emotional
This is why body oder changes for exercise vs stress
If you wink you can pick up GSR on your finger tips
But don't get that on the arm
Body media prototype that does EKG from the arm band
GSR is active and actually draws crud to the skin - early versions really bad for this especially for sweaty people. Switch current direction to help.
Accelerometer for context sensing. But a lot of activities look similar! Stairs vs walking. GSR used in energy expenditure algorithms, but actually most useful in context sensing. Eg driving - motion but no sweating.
GSR does saturate - you get too sweaty, and information, even changes become noisy and meaningless.
Military how are you test - how many fingers? Can't we find sensors for this? But turns out pulse breathing and lots of stuff stays far too normal during bad trauma. Shock index. But they were able to predict 90% of time the right level of shock index using the arm band, in a clinical setting.
Using band plus info about what they ate, they are trying to get blood glucose levels. Sort of working.
GSR plus heat flux is super interesting during sleep - your thermal regulation really changes. With heart rate, maybe we will be able to sleep stage. Sleep apnea lowers the sleep efficiency measure that we have now.
43 channels of data after some basic analysis. 32 Hz is the fastest data reporting speed. 128 Hz for the prototype with heart rate data.
Most problematic confounds - exercise machines we have not studied. New strange motions.
ECG on arm fails if you use your arm muscles. Brushing teeth - looks like a heart attack.
Wrist is only slightly better than arm for GSR
Yoga is a hard case - if you sweat, works, otherwise not so much.
They have a rich annotated data set from employees and friends.
They have looked at stress, but not terribly much in the product
Free living databases - huge opportunity going forward.
Ambient light sensor - how many calories are you burning? Where on the body would you put it. Interview question that they actually use!
Inductance sensors - apparently you can sense breathing rate.
Microphone only worked for them on the neck - everywhere else they just got a lot of noise.
They have contract from the military to do noninvasive glucose sensing
- week long baseline using continuous glucose monitor, then a week of their sensor, get ok data - but only as long as their medical condition doesn't change. You have to recalibrate every 2 to 3 months.
He too sees that the middle wear piece is missing but super important
One of the most powerful things that can happen is when some else is looking at your data.
Email Dave Andre for access to the beta dandre@body media.com or davidandre.com
GSR sensor is also on off switch
It's been used on horses and dogs
BodyMedia has 54 employees - this is the third time they have been big, dot com, clinical, now consumer. Founded mid nineties.
Wavelet analysis for heart rate variability
Email Dave about Kevin Warn about blood replacement.
Weavrs.com are the augmentation system for quantified self data.
What color is hungry - or any other mood. Rust brown. People choose colors and we aggregate
Consensus ontology for mood words
Like a crowd sourced PANAS
Xperimeent engagement model - phases of how people feel about the experiment as it progresses
People are not so good at remembering how they have been - and also doctors only give 20 seconds for the answer to that question
2001 the pain clicker
2002 the pain scanner - bar codes in a booklet
Debugging your code through you - biometrics as you code.
Kyle Machius open you.org
Fred trotter the programmable self open source tools for self tracking and behavior change
Hacktivist - I code for social change
Cautious patient foundation - patient safety - eg concern with medical errors
Fogg behavior model
Against gamification for health - AA is gamified - levels and badges, weight watchers points, etc. Gamification of exercise is sports - so this stuff is already being done! But needs to be done more on the providers side. Obamacare is gonna change things - maybe we can gamify pro ciders side.
Surgical checklist is single greatest impact in recent decades
Elephant and rider analogy/model. Your job as the rider is to control the elephant. Shape the path. He doesn't like the model for mental health. If your analogies are bad that feeds into the problems themselves.
Herd of buffalo - we have lots of different instincts. Smoking and weight gain.
Losing money is one of the buffalos you can use to control the herd - stickk
Quit drinking by taking up smoking
Dog training, super nanny - very good behavior modification stuff
Shock collar - only when dogs life is in danger, and shock myself twice for every time I shock the dog
Programmable self - shock system on my arm - gonna test it on himself
Doesn't think it's gonna work - physical punishment doesn't work in the same way as psychological retraining stuff
Shock himself based on failure to do things - not automatic, read failure and self shock.
Pacting - swap sex of money for something.
There is a strong movement to use Twitter as the backend for quantified stuff
Posting withings weight to Twitter & facebook is the only quant hack he's done that actually worked for him - he's lost 20 pounds, even though nobody else cares when they see his posts.
Behavior change stuff is very subtle
When I go to a conference I gain about 5 pounds.
Join the programmable self mailing list
Book sticks and carrots is a must read
You are not one entity in time - you need to plan for your future self. Bind your future self like Ulysses and the sirens
Status presentation using timed release of tweets
Social relationships matter to people at a much lower level than we think, ie it's a buffalo, maybe many of them... That high school girl I had a crush on doesn't even notice, but he thinks about her and others when those tweets go out
For the most part behavior change does not work
It's about what works for you
There's very few fundamental motivations that we can play with from a technological standpoint.
Stage models of behavior change - do easy things first.
He needs three things: testers, coders, feedback from scientists
Solutions vs tools
Stickk - if you spent $5000 you will lie! If you spend $1, you won't care and won't get behavior change - how do you find the sweet spot.
Measuring whether you achieved your goal is way more important than the fine grained detail.
Emailing yourself photo of fat guy when data shows you are missing your goals.
Sex is way underutilized as a motivational thing. It's rewarded exercise.
Cyberdoms as feedback sex reward workers
- doing extreme quantification of companies.
Problem with a lot of data is over fitting
Control for season - people are less active in the winter
Ultimate goal: optimal health
When someone buys a device they expect it to do something immediately - but a lot of data is really only useful once you have years of data.
Extreme quantification to measure your performance against the best possible performance - emg shorts to really close the loop on doing the best possible workout
When you have too little data it's too easy to find patterns that are not there
RescueTime - downloadable apps tracks what software websites you are using.
Article makers and managers
Last.fm does really good music tracking stuff.
Mediawiki can do semantic analysis of the stuff that's in it
Gravity.com - watch how bodies of text change over time, do an interest graph
Qualitative research software - google that term
Airport scanners - how does my posture change over time
Toilets - huge untapped opportunity
Color - location tracking using cell phone carrier information.
Gratitude tracking. Martin Seligman
Upload your notes to qswiki.com
Quantified self is part of a larger trend
What does technology want us to do?
PC: File folder desktop
Web: Page link web
Web 2.0: Stream tag cloud
Experience - the art of shifting attention. Balance between paying attention to the data vs wanting to not pay attention - having it collected automatically
Sharing: any data that can be shared will be shared.
Quantified self = intimate surveillance? Permission
Surveillance leads to personalization, the two must go hand in hand, private = generic, transparent = personalized
Shifting from ownership to access
The scientific method will change more in the next 50 years than in the past 400 years.
- keep negative results
- triple blind - you don't even know you're doing an experiment
Qs is part of the next stage of the scientific method
Subscribe to Posts [Atom]