I wrote previously on how Data Science techniques for optimizing product growth might have unintended consequences. A product becoming more addictive or surfacing more divisive content are typical examples. Here I explore a different type of example.
When I ask Google Maps how to walk from one location in the city to another, there are many possible routes it may recommend. Some routes may better develop my ability to navigate in the city; for example, a route staying on a few major roads may help me get to know the structure of the city, while a route that makes many turns on many small streets may be too confusing to contribute to my understanding of the shape of the city.
Google presumably runs experiments, testing different methods of selecting routes. As they develop new methods, they naturally want to test these methods to see which are best, both in terms of user experience and in terms of the success of the product.
It’s likely that these experiments are analyzed using growth metrics – seeing which methods lead to the greatest use of the product. Generally, we imagine that this measures both the success of the product (more use = more revenue) and the user experience (if people are using it more, it must be because they’re happier with it). However, what we measure (how much people use the product) and what we want to measure (how good the user experience of the product is) are not quite the same thing.
It’s possible that I may use Google Maps less often over time (or stop using it entirely) if I develop a strong sense of direction in the city I live in. This means that when Google tests different methods, they could find that those methods that lead to more confusing routes, and thus to less development of the user’s sense of direction, actually lead to greater product growth, and thus be selected for use.
Without any intention to do so, the naive optimization of Google Maps for product growth could cause the product to “create learned helplessness” and interfere with the users’ ability to navigate on their own.
Ultimately, I do not believe that Google would intentionally sabotage my sense of direction in order to increase my dependence on their products. I don’t even think it’s likely that the effect I write about could realistically happen, unintentionally or intentionally. I, do however, think it’s an interesting examination of the danger in blindly optimizing products for growth.
Wrong, but useful 