Philosophy, Running, Gaming & the Quantified Self
“The unquantified life is not worth living.”
While the idea of quantifying one’s life is an old idea, one growing tech trend is the use of devices and apps to quantify the self. As a runner, I started quantifying my running life back in 1987—that is when I started keeping a daily running log. Back then, the smartest wearable was probably a Casio calculator watch, so I kept all my records on paper. In fact, I still do—as a matter of tradition.
I use my running log to track my distance, running route, time, conditions, how I felt during the run, the number of time I have run in the shoes and other data I feel like noting at the time. I also keep a race log and a log of my yearly mileage. So, like Ben Franklin, I was quantifying before it became cool. Like Ben, I have found this rather useful—looking at my records allows me to form hypotheses regarding what factors contribute to injury (high mileage, hill work and lots of racing) and what results in better race times (rest and speed work). As such, I am sold on the value of quantification—at least in running.
In addition to my ORD (Obsessive Running/Racing Disorder) I am also a nerdcore gamer—I started with the original D&D basic set and still have shelves (and now hard drive space) devoted to games. In the sort of games I play the most, such as Pathfinder, Call of Cthulu and World of Warcraft the characters are fully quantified. That is, the character is a set of stats such as strength, constitution, dexterity, hit points, and sanity. Such games also feature sets of rules for the effects of the numbers as well as clear optimization paths. Given this background in gaming, it is not surprising that I see the quantified self as an attempt by a person to create, in effect, a character sheet for herself. That is, to see all her stats and to look for ways to optimize this character that is a model of the self. As such, I get the appeal. Naturally, as a philosopher I do have some concerns about the quantified self and how that relates to the qualities of life—but that is a matter for another time. For now, I will focus on a brief critical look at the quantified self.
Two obvious concerns about the quantified data regarding the self (or whatever is being measured) are questions regarding the accuracy of the data and questions regarding the usefulness of the data. To use an obvious example about accuracy, there is the question of how well a wearable really measures sleep. In regards to usefulness, I wonder what I would garner from knowing how long I chew my food or the frequency of my urination.
The accuracy of the data is primarily a technical or engineering problem. As such, accuracy problems can be addressed with improvements in the hardware and software. Of course, until the data is known to be reasonably accurate, then it should be regarded with due skepticism.
The usefulness of the data is partially a subjective matter. That is, what counts as useful data will vary from person to person based on their needs and goals. For example, knowing how many steps I have taken at work is probably not useful data for me—since I run about 60 miles per week, that little amount of walking is most likely insignificant in regards to my fitness. However, someone who has no other exercise might find such data very useful. As might be suspected, it is easy to be buried under an avalanche of data and a serious challenge for anyone who wants to make use of the slew of apps and devices is to sort out the data that would actually be useful from the thousands or millions of data bits that would not be useful.
Another area of obvious concern is the reasoning applied to the data. Some devices and apps supply raw data, such as miles run or average heartrate. Others purport to offer an analysis of the data—that is, to engage in automated reasoning regarding the data. In any case, the user will need to engage in some form of reasoning to use the data.
In philosophy, the two main basic tools in regards to personal causal reasoning are derived from Mill’s classic methods. One method is commonly known as the method of agreement (or common thread reasoning). Using this method involves considering an effect (such as poor sleep or a knee injury) that has occurred multiple times (at least twice). The basic idea is to consider the factor or factors that are present each time the effect occurs and to sort through them to find the likely cause (or causes). For example, a runner might find that all her knee issues follow times when she takes up extensive hill work, thus suggesting the hill work as a causal factor.
The second method is commonly known as the method of difference. Using this method requires at least two situations: one in which the effect in question has occurred and one in which it has not. The reasoning process involves considering the differences between the two situations and sorting out which factor (or factors) is the likely cause. For example, a runner might find that when he does well in a race, he always gets plenty of rest the week before. When he does poorly, he is always poorly rested due to lack of sleep. This would indicate that there is a connection between the rest and race performance.
There are, of course, many classic causal fallacies that serve as traps for such reasoning. One of the best known is post hoc, ergo propter hoc (after this, therefore because of this). This fallacy occurs when it is inferred that A causes B simply because A is followed by B. For example, a person might note that her device showed that she walked more stairs during the week before doing well at a 5K and simply infer that walking more stairs caused her to run better. There could be a connection, but it would take more evidence to support that conclusion.
Other causal reasoning errors include the aptly named ignoring a common cause (thinking that A must cause B without considering that A and B might both be the effects of C), ignoring the possibility of coincidence (thinking A causes B without considering that it is merely coincidence) and reversing causation (taking A to cause B without considering that B might have caused A). There are, of course, the various sayings that warn about poor causal thinking, such as “correlation is not causation” and these tend to correlate with named errors in causal reasoning.
People obviously vary in their ability to engage in causal reasoning and this would also apply to the design of the various apps and devices that purport to inform their users about the data they gather. Obviously, the better a person is at philosophical (in this case causal) reasoning, the better she will be able to use the data.
The takeaway, then, is that there are at least three important considerations regarding the quantification of the self in regards to the data. These are the accuracy of the data, the usefulness of the data, and the quality of the reasoning (be it automated or done by the person) applied to the data.