Don’t Make Me Think, But Don’t Prevent Me From Thinking

Yesterday morning I had to make a series of choices on how I was going to get to work. As a consultant, each day can be quite different in this matter. Yesterday, I had a noon time meeting with a client out in Woodside, NJ which can be about a 2 hr drive from my apartment in Brooklyn. Typically, I would ride with my business partner, but that morning his car wasn’t available so I was driving and we would meet in the city. My fiance also had to get to work in the city about 30 minutes before I met with my business partner and there I was faced with a new set of variables to make an important choice about my path to work.

Will my fiance make it work on time if I drop her on the way? What route should I take?

So as anyone else would do, I looked at the different map apps. I looked at Waze and I looked at Google Maps. Waze originally said the drive would be 1hr and 15 minutes, which made sense and Google gave us a similar estimate. I thought, great, we have enough time to get out the apartment and in the car to make it on time. When we got to the car, we were running a bit behind schedule so we considered perhaps just having her take a subway ride instead. I looked the Waze app and it said 1hr 6 minutes and Google Maps said 1 hr.

Typically, for a local drive, I trust Waze. Waze seems to have a much better sense of off highway streets which can really save time on a local ride. As we got in the car and started following the route, we noticed Waze had us get on the BQE instead crossing the bridge straight from Brooklyn to Manhattan and we adjusted our route based on our own understanding of NYC traffic. Waze made no time adjustment as we did this, so we took it to mean there was no expected change.

Well, we followed the route all the way through and as we got to the bridge, traffic was bumper-to-bumper with no movement and ultimately, she was 20 minutes late to work on a trip that took about 1hr and 30 minutes. Along the path, Waze began adjusting the time adding minutes at each delay. Afterwards, I never really believed that there was ever a path that would have gotten us to her office in the 1 hour and 6 minutes as predicted by the app.

After this trip, I was left wondering, what went wrong? How did Waze calculate our ride? Why was it off? Was it my driving? I don’t know the answer and without thorough research, I will never know. Previously, my only regular experience of this frustration has been with the weather man. That ever doubtful and prone-to-error prediction of what the weather will be tomorrow. My decision making was thrown into the black magic of automation and algorithms. Someone out there knows why this happened and with enough searching on Google, I am certain to find an answer, but should that be necessary?

Automation processes and great user experience design have done such a great job of reducing the decision making process for us and, when done right, they make smarter decisions than we would make ourselves. It is amazing as to how intuitive Google search is today at understanding what we are searching for and prioritizing results for best matches. It’s been years since I’ve gone beyond even 2 pages of a Google search to find what I’m actually looking for. Yet there is a growing a gap in our knowledge of the decision-making process that is happening on the behalf of our own minds. With the accelerated advancements in machine learning, the gap is only bound to grow at an exponential level and this presents a critical challenge to address as UX designers.

Personally, I have no idea of the rudimentary levels of decision making that happened on my path to work. Was the timing off because it didn’t calculate timing of turning from one street on to another? Maybe the algorithm went by an average movement in time of streets over a distance that didn’t account for the difference in time it takes to clear one segment of the street vs. the street as a whole. Maybe there weren’t enough drivers contributing to the route time that morning to give an accurate reading it was going on an average over a year or so?

Why can’t Waze intelligently tell me that the timing may be off based on the fact that today it’s working off of 100 data entries vs. an average of 500 and that it’s relying heavily on past data to calculate my route today?

We are constantly being trained to relate to our automated devices as magical, all-knowing entities and the only time we learn that they are not all knowing or magical is when it’s too late, if ever. Today, I think we are still forgiving of mistakes. I have no intention to stop using Waze and I am very grateful that it exists. But I don’t expect that this level of algorithm ambiguity will always be acceptable. I want to know why Spotify populates the songs it does on my Discover weekly list. I still never really understood the specificity of Pandora’s Music Genome Project.

We used to make decisions on the music we want to listen to and the route we want to take to work and we began to accept that algorithms can do it better than we can on our own. Accept when they don’t come through! The ability to anticipate potential failure is taken a way from us. Therefore, algorithms in the dark are no longer acceptable for providing a great user experience.

Algorithm weaknesses should be presented to us when they come up. Instead of having an option simply to say I like this song or don’t like this song, I’d like an option telling me why you selected it in the first place and I’d like to provide input as to what I want you to do differently.

I believe advancements in machine learning technology can either lead us further down the rabbit hole of making decisions for us or augment our decision making process. I clearly advocate the latter. Our current path of exponential growth in technological capacities will continue be slowed down as they hit the market if they don’t start addressing this. If I experienced frustration for not being able to smartly gauge the time it will take to get to work, I can’t imagine the frustration of a driverless car taking me to work and making loads of micro-decisions without any explanation.