Memory Rank: Bias or Heuristic?

2022-03-10: I’m starting to wonder if the “availability heuristic” reflects only a portion of the “[[/_paper/why-forget|fluency-heuristic]]”.

The word “bias” conjures up all sorts of negative feelings and nudges us toward imagery of people are faulty biological meat-bags with pugnacious monkey minds. We think of a stubborn colleague or some stranger on the street that was just impossible.

Many have heard of the “availability bias”: It is the mechanism by which our minds produce “immediate examples…when evaluating a specific topic, concept, method, or decision.” The concept comes from Decision Theory and Behavioral Economics. There, the original term was “availability heuristic”.

“Heuristic” doesn’t bring up the same imagery. At least not for me. Heuristics are cold, hard, numbers calculated by computers. They are quick solutions thrown together to save time, energy, and maybe money. They are not optimized and so come with addendums and the expected grain of salt to consider…But they are valid tools and are useful in certain scenarios.

When I think about the underlying mechanics of availability, I tend to think “heuristic” comes closer to the mark. But this begs the question: When does a useful “heuristic” become a harmful “bias”?

Reinforcing Reason 🐝

I really enjoyed reading rationality. Besides its strengths as a manual for rationality, reason, and critical thinking, Pinker occasionally makes off-hand comments that are stunningly insightful and deserve to have so much more said about them. Pinker could not possibly expound upon every single one of these ideas or else the book might be 1000s of pages long…That said, I felt compelled to embellish on one tidbit in particular:

“We use the ranking from our brain’s search engine…as our best guess of…probabilities.”

~ Steven Pinker, rationality (Emphasis added)

Let’s unpack that…📦

Page Rank 🐝

One particularly famous algorithm is called Page Rank. This algorithm was the original driving force behind Google’s search engine, which answers our questions and finds relevant web pages. Search has evolved much since the algorithm’s first inception, but its basic function works something like this:

A search engine will rank websites and pages by their relative “importance”. “Importance” is calculated by counting the number of backlinks each website has. If website A links to website B, then website B has one backlink. If websites A, B, and C link to website D then website D has three backlinks. And so on.

Needless to say, this algorithm was a breakthrough in pointing people to relevant pages quickly and it has worked remarkably well over the years. I think it’s safe to say that everyone would agree this algorithm has been, at a minimum, useful.

Memory Rank 🐝

Let’s zoom out on our quote a bit:

“To estimate risk, we should tally the number of instances of an event and mentally divide it by the number of occasions on which it could have taken place.

“Yet one of the signature findings in the science of human judgment is that this is not how human probability estimation generally works…

“…people judge the probability of events by the ease with which instances come into mind…We use the ranking from our brain’s search engine…as our best guess of the probabilities. The heuristic exploits a feature of human memory, namely that recall is affected by frequency: the more often we encounter something, the stronger the trace it leaves in our brains. So working backwards and estimating frequency from recallability often works serviceably well.”

~ Steven Pinker, rationality

Pinker is referring to the forgetting curve (implicitly) and the availability heuristic (explicitly). Each factoid we come across undergoes the process of forgetting at a rate dictated by the forgetting curve (which is an exponential decay) and the more we come across that factoid the more easily we remember it and the more available that factoid becomes to us. We then, in turn, derive likely probabilities from this sense of ease or availability.

It’s almost as if your brain is organically managing its own page rank algorithm and proferring up results based on the number of “links”, or number of times you’ve experienced the factoid. Since each encounter leaves a stronger mental trace, no actual counting is necessary. We just feel and experience it. It just sort of happens.

Bias In The System 🐝

As we all know, Google and other search engines have served us remarkably well. But we have also learned from search engines that “number of backlinks” can fall short or be gamed: JCPenny once gamed the Google search algorithm by creating many links that referenced their own site and products so that they would rank higher in the search results.1 Similarly, our brains work a certain way and that certain way can also be gamed or biased: As media and news outlets mention more and more specific types of incidents, those incidents seem more prevalent to us.

And these aren’t the only hacks that can skew result rankings. If Google search only returned websites or pages that related to politics it would be rendered a (mostly) useless tool for scientific endeavors. If search only returned urls for websites in English, all other languages would be undiscoverable. And similarly, as more and more of our everyday interactions are sorted into like-minded buckets, and less and less random or serendipitous chance bubbles throughout our lives, our built-in systems will tend to be biased in certain directions.2 If we don’t interact with different perspectives than ours we will never be made aware of the vast complexities of the world outside our own purview.

Even well-functioning systems can be gamed, hacked, or otherwise manipulated. And it seems clear when the aforementioned systems that measure importance or relevance reside in an environment that feed it biased information, the system itself cannot help but be biased in a similar manner.

Bias in, bias out.3

  1. There’s probably a parallel lesson to be learned from citations in academia somewhere. 

  2. There’s already been a lot of talk around [[echo-chamber]]s, [[the-big-sort|self-sorting]], and too much homogeneity of thought in society, so maybe I’m beating a dead horse here… 

  3. Obligatory explicit acknowledgement that the “availability heuristic” is still a heuristic and that it generally does not outperform hard (big) data and accurate statistical analysis.