ShareThis1 |
I'll put together a 10 minute PowerPoint presentation of TTL ideas as they relate to my "Experiment-driven life" mini-manifesto (http://www.matthewcornell.org/2010/06/the-experiment-driven-life.html) at the Boston Quantified Self #3 meetup on Tue 2010-08-10, which gives me about three days total to prepare. The event is here: http://www.meetup.com/bostonQS/calendar/14132216/ |
2 |
Pulling together a presentation that's Good Enough, then presenting it, is #1. The main test, though, is whether I can convey some of the TTL ideas to the audience. If I get questions during or after I'll mark it as a home run. Being invited somewhere else to speak would be great too. |
3 |
Tue is the hard deadline. |
4 |
This is a challenging experiment for me - I'm a perfectionist; creating presentations, even 10 minute ones, is like pulling teeth; public speaking is scary (though once I practice I'm happy doing it); putting something I've created out into the world feels personal; this feels big and important, so it hangs over me all the time, even when doing something else; etc. To enjoy it:
|
Tags Think, Try, Learn, meta, presenting, public speaking, quantified self, powerpoint
Comments & Observations
Matthew Cornell It went very well. I congratulate myself on my first my first time sharing the work in a public forum. Well done! Some notes:
* On putting the show together:
o It is *hard* to make a great 10 minute presentation, i.e., one that stimulates, is humorous, moves quickly, is draws the audience in using multiple modalities, doesn't have bullet points, and gives them a solid take-awy.
o I used Lessig-style text (large white typewriter text, only a few words, on a black background) plus lots of images.
o Picking good images is fun, but takes thinking (stock photography doesn't hack it).
o My rule for public speaking is to never go over, and ideally use less time than allocated. To keep it to 10 minutes I had to slay some sacred cows (my lobster slide, and my Dorothy Parker quote)
o One way I got kept it down was to TALK LESS!
o Interestingly, the ratio of time spent to presentation time was 60:1 (one hour for every one minute of the presentation).
In the end I was (and still am) very happy with what I created. You can watch the finished work here, BTW: http://www.slideshare.net/matthewcornell/quantified-selfexperimentdrivenlifenoanimation-5020668
* On the presentation itself:
o Lesson learned: I should have brought a list of my current experiments. I blanked when asked this obvious question. Related lesson is to brainstorm the hard questions I might be asked.
o I had a handful of questions at the Q&A after the talk, but only one person came up afterwards, and that was to share a site he thought I'd like. I expected more interest.
o Surprise: I had to give the talk with my laptop on the opposite side from which I'd practice. Thinking about it beforehand threw me off a bit, but it was fine once I got into it.
o There were definitely some of "my people" here, but they seemed less interested in the philosophy than the applications and technical details. Maybe it's just the audience, but still - stories are so important. This gets the the "why track data" point of the presentation.
o Surprise: only one woman out of 20 total!
o I don't think anyone signed up for Edison.
The ultimate question is whether it was worth it, and should I do something similar again? While driving home I had a letdown, but it's a week later and I feel extremely happy having done it. Of course the YNK factor is important - what will stick or, better, bounce higher as a result?
Aug 20, 2010
Will Ware I didn't get to the meetup but I stumbled across your slideshow via a random Facebook connection. I wasn't aware there was a QS meetup in Boston. Edison looks cool and you have now scored at least one sign-up.
If I can presume to make a suggestion, you might create ways on the website for people to maintain numerical data that could be selectively shared with other users, turned into 2d and 3d graphs, etc. If I have permission to see somebody's data for a particular experiment, there should be an API allowing me to pull the data into a scripting language, and hopefully some kind of semantic markup saying what the data is, and what the units are.
Aug 22, 2010
Matthew Cornell I'm really happy to have you here, Will. I completely agree re: the quantitative data layer. It's essential for the TTL computational engine that's in version 2. I'm open to partnering with someone, but I'm not at all convinced that it's the right way to go. There are many tools to capture personal data, but I haven't been blown away by any yet. Also, as I try to lay out in the talk, they need to be in the context of an experiment to be most useful. Right now you have to use an external tool (some use Google Docs) and link to them from edison.
The API is a brilliant idea. I've now been hit on the head enough times to pay attention. Question: How would you envision using it?
I agree that sharing needs a finer granularity. We've given it no attention until now, but a trusted circle makes a lot of sense. I've received inquiries from folks like psychologists who want to use it for patients' adopting of suggested behavioral modification experiments.
Aug 23, 2010
Will Ware I think a really well-designed API would be vastly more useful and flexible than a good-enough one, and any quick specific suggestion would be of the good-enough variety. Here are a few principles I think are good. I'll cogitate over the next few days and try to come up with something more specific.
(1) There should be huge flexibility in how data is structured. It should be very easy to set up a uniform structure (like a SQL schema) but it should be possible with a little more work to have some much more complex nuanced structure.
(2) The structure of the data needs to be discoverable by any program or script with permission to access the data. A few different people have looked at trade-offs between learnability, useability, and discoverability (http://haacked.com/archive/2008/11/06/usability-vs-discoverability.aspx).
(3) There should be semantic markup: what the units are, what conditions existed when the data was collected, any context info needed to correctly interpret or understand the data.
(4) Data should be human-readable, or it should be easy to render it in human-readable form.
I occasionally attend semantic web meetups where they talk a lot about linked data (http://en.wikipedia.org/wiki/Linked_Data) and I believe those guys are definitely onto something.
As far as granularity of sharing, it probably makes sense for each user to be able to create groups of users and to use boolean combinations (maybe just set unions?) of these groups (and of single users) to form the group of people allowed to look at a particular data set.
Aug 24, 2010
Matthew Cornell OK, thanks very much for that Will. I've captured it in the project's To Do's.
Aug 24, 2010
Will Ware I gave this a little more thought and started to put something together. I can see a few places to apply these ideas so I'm going to try to find more time to develop this a little more. Here's what I've got so far.
http://snipplr.com/view/39690/puttering-with-a-proposal-for-a-web-api-for-numerical-data-sharing/
I didn't do a lot with data formatting yet, but I tried not to make any early decisions that would paint it into a corner.
Aug 29, 2010
Matthew Cornell Wow! I'll give it a serious look, and I'll ask our (very) part-time developer to look at it. I'm forging ahead figuring out how the quantitative data layer will work, and your ideas will be helpful. Thanks so much!
Aug 29, 2010
Matthew Cornell Having created the slidecast, put it up, then having Alex at the QuantifiedSelf blog mention it has continued to pay off. I spoke with two leaders who are creating software that has elements of what we're thinking up with edison. Very exciting to hear I'm on the right path!
Sep 09, 2010
Will Ware Matthew, are you going to the Life Extension conference next month in Palo Alto? I'm planning to go, and maybe see about putting together a BoF session about data sharing in QS if there is interest.
Sep 11, 2010
Matthew Cornell Hi Will. First I heard of it. The BoF session would be fun. Tell me: What do you hope to get out of the conference? Maybe I'll go to something similar, or next year. Thanks for letting me know!
Sep 11, 2010
Will Ware Sorry to hear you won't be able to make it this year. It's put on by the Foresight Institute, most of whose conferences are in nanotechnology (same people are interested in life extension, and in QS) and which I've been attending on and off for years. So partly I hope to run into interesting people I don't get to see very often. Hoping also to get a kick in the pants to eat healthier, get more exercise, hear interesting and hopeful news, etc. In recent years I've gotten keenly interested in automating some or all of the scientific process (http://bit.ly/9S19ch) hoping to rapidly advance medical progress, and talking to QS hackers about sharing data and semantically marking it up to be processed by scripts dovetails nicely with that.
Sep 11, 2010
Matthew Cornell Re: automating science, I reviewed your post, but I didn't get it. Deserves more study. My focus is in individuals doing self-experiments, and my thinking is the datasets are not large, but they could be complex.
Reminds me that I recently came across tools for experiment workflow when I was searching for inspiring flowcharts for a "Try" one I put together. I wonder how the features in those tools relate to your ideas. (For example, http://research.microsoft.com/en-us/collaboration/tools/trident.aspx)
Sep 16, 2010
Will Ware Now that I think about it more, the things I was thinking about are probably overkill for QS, and would involve more setup effort than would make sense. The Trident thing looks interesting. My impression from the phrase "Semantic tagging to enable a researcher to find a service based on what it does, or is meant to do, and what it consumes as inputs and produces as outputs" is that they have computational blocks that can flag certain classes of errors if you feed them inappropriate data (e.g. fluid volume going into a block that translates Fahrenheit to Centigrade) and correctly tags their outputs for further error-checking. Kinda similar to what I'm thinking about but it looks like assembling the blocks into a bigger process is always a manual activity.
Sep 27, 2010
Matthew Cornell Thanks for giving it your brainpower, Will. I thought about it too and I agree that something pretty simple should do the trick, at least for the uses I have in mind.
Sep 29, 2010