This week was somewhat easier in terms of familiarity with the subject material–COMP683 was just last term, so the subject of analytics is still very fresh in my mind. I’m not sure if I’ve understood the distinction between intrinsic and extrinsic purposes correctly, but in my mind, “intrinsic analytics” are how the system knows what it needs to do its job, while “extrinsic analytics” are how we learn about how the system is being used and how it could be further usefully modified.
I’ve always liked thinking about how humans run their own algorithms. To me, it’s all about patterns. Sometimes, those patterns can comfort us or provide a template for how to deal with unfamiliar situations. Sometimes, those patterns can be subverted to make a point or trigger deeper examination. With different starting conditions and varied inputs during execution, different patterns of behavior are selected. People hate being compared to a machine that acts by rote, but the truth is that we all have our habits and a general “common sense” understanding of how we should comport ourselves. (I note that “common sense” is neither common nor especially sensible! A lot of it depends on your social position in life, your physical community, and the particular situation at hand. People in the same physical place, but in a different situation or with a different background, may always have a different “common sense” than us, but it is no less valid because of it!)
I had not encountered the term “Matthew effect,” but I am of course familiar with various bandwagon effects. For example, in a busy registration line at a conference! One of the most important things to do to ensure a good experience for the attendees is to minimize the line. One solid way to do this is to actually secretly open some hours in advance of the public time, in order to get people registered as soon as they start trying to line up. If people see a line forming, their first instinct is “Oh, I should grab a place in the line now or I’ll have to wait longer later!” With that, the more people line up, the more people will keep lining up, and suddenly you have a registration line that is six hours long. Instead, if you process people immediately, everybody passing by sees a minimal line. They don’t feel the need to line up in fear of encountering a longer wait time later, and thus a lineup can be kept manageable and quick to process, and everybody has a much better experience overall.
And that’s an example of a human algorithm. Understanding how people in general tend to behave allows us to design systems to take advantage of (or stymie) those behaviors. It’s not perfect, being based on probability, but on balance it’s incredibly useful.
In fact, that was one of my biggest insights this week. As much as we might blame “the machine” for constructing echo chambers and filter bubbles around us… we humans are probably just as much to blame, simply by selecting options that are comfortable versus more challenging. I can’t fault us for that, but it’s important to note that even if we built a perfectly unbiased Web, our own human behavior might reconstruct the echo chambers around us. In that respect, then, perhaps we do need to construct our systems with a little bit of bias–to help us push outward and continue challenging ourselves! That wouldn’t be to maximize engagement, necessarily; it would have to be built for a different set of rationales altogether.
As a final aside in this reflection, I suppose it’s reasonable to note that I wish I had more time in the week. There’s so much I want to read but simply do not have time for… And I feel bad, because there’s a huge list and even if I only select the items I’m keenest on reading, I still don’t have enough time to read them all to the depth I want. It’s quite frustrating, even disappointing.
Leave a Reply