We've recently begun to discuss how big business uses feedback from its users to impact how the business suggests new products to the user. Examples have included sites such as Amazon and Pandora.
One of the topics we discussed in class had special interest to me - unused data. As discussed, users are being presented with a set of "options". Users of Pandora have a seemingly binary set of options "like" and "dislike". Amazon users have a set of suggested products that vary in size, shape, type, and color. However, there is one option that has gone overlooked - the user's inaction.
One can consider these algorithms as machine learning in that the user is training the algorithm by active selection to provide better options for suggestion. But, what's missing is an active address to when the user doesn't respond. This "inaction" could prove to be very valuable data.
I recently watched a TED talk called "Shawn Achor: The happy secret to better work". Mr. Achor is a psychologist and counselor that proposes a paradigm shift in how we analyze normality. Shawn shows a data set in the beginning that has an obvious trend and one very random outlier. He goes on to say how statistics trains us to go ahead and delete that data so that we can fit our set with a line of best fit and move on with our analysis. He then proposes a question - why? Why is it that we see no value in this data point? It is a real event that took place. Sure, there are obvious reasons that we don't look at it - cost, time, effort, etc. But, what if there were underlying factors in the set of outliers that we continually "delete".
Mr. Achor goes on to suggest that happiness has developed a narrow image. People view success as analogous to happiness. You can argue with that point all you want, but I think we can all agree on this being a norm. The problem with this norm, as Shawn points out, is that it's completely illogical. If we view happiness to be on the other side of a milestone, then as soon as we achieve that milestone we develop yet another and happiness comes to reside on the other side of our new milestone. Thus, we are effectively acting as the proverbial horse with a carrot attached to a stick... in a continual chase.
Mr. Achor then suggests what his research has found concerning this paradigm shift. If you're interested, you can watch the video here. He's very funny and this is one of my favorite talks. I highly recommend you check it out.
So, how does this relate? What Shawn points out is something that I think these sites are missing as well. If we only concern ourselves with one aspect of the picture, we never see the entire thing. What I'm suggesting is for Pandora (for example) to monitor when songs don't receive a "like" or "dislike" when the algorithm suggested a song that should have gotten one or the other and then ask "why?" Or, when a song that already has a "like" gets skipped. Is there a trend? I believe there absolutely would be. I know that my song selection depends on so many other things than just my taste. Environment, mood, time of day, and other things place into my selection as well.
This goes for Amazon as well. The only times I'm buying textbooks are at the beginning of semesters. They should know that by now. Thus, I should be getting discounts during those times and prompts at the end of semesters to sell back. These types of buying habits should make sense because I have not opened a single email from them to look at other "suggested engineering books". My action to not select and read these emails is very insightful data and should add to their approach
No comments:
Post a Comment