This is a summary of [Eliezer’s article on Availability bias][ele_avai].

Availabilty

Availbility hueristic is one way in which we humans universally operate. We tend to judge the probability of an event, based on which examples of the event come to mind.

A study by Lichtenstein et al. in 1978, that judged the ‘frequency of lethal events’, informed that subjects thought accidents to be more frequent than disease; and suicide to be more frequent than homicide. This judgement was quite wrong as the actual data suggested that diseases killed 16 times more than accidents; And homicides two times more than suicides.

Our judgement in determining the frequency of future events sucks. Got it! Obvious hypotheses could be that suicides and accidents are more dramatic, that we tend to remember them well; or that they get featured in the tabloids and the media much more often, resulting in selective availability.

In 1979, a followup study by Combs and Slovic showed that the skewed probability judgements correlated strongly (.85 and .89) with skewed reporting frequencies in two newspapers. This doesn’t disentangle whether murders are more available to memory because they are more reported-on, or whether newspapers report more on murders because murders are more vivid (hence also more remembered). But either way, an availability bias is at work.

It is clear that in order to conclude that ‘newspapers report more on murders because murders are more vivid’, we need more information. However, regarding whether murders are more available to memory because they are more reported-on, seems slightly tricky as I find myself spending quite some time on it, without fruit. I currently don’t see the difference between the ‘followup study by Combs and Slovic’ while comparing it to the statement made immediately after.

Investigation into the not-clear

Probability judgements are strongly correlated to reporting frequencies. I believe Eleizer added ‘skewed’ to inform that our judgments are wrong af! but why did he add ‘skewed’ to the reporting? Probably because reporters were assumed to be selectively reporting, which is not ‘Truth’, hence skewed. Probability judgements(sentence 1) I assume are a result of availability(sentence 2), i.e. information available to memory allowing us to make guesses. Reporting frequencies(sentence 1) seem to be the same as referring to more reported-on. And Sentence 1 and 2 seem to be practically the same.

Here are my following hypothesis where the error might be in my understanding:

First hypothesis is that the assumption ‘Probability judgements(sentence 1) I assume are a result of availability(sentence 2)’ is wrong. Second is based on the fact that they use the term ‘more’; That Eleizer is questioning if more reports => more availability to the memory. Third- Correctional data might not actually mean anything, as evidenced in sentence 2.

Hypothesis 1 seems not so probable. More reports needn’t mean more availability to the memory; for example, reports might not be proportional to availability, that is to say that it could be an exponential or an x^2 graph; But as I try to explain what the other examples could be to support hypothesis 2, I am unable to come up with anything. Thats where my doubts regarding this hypothesis still continue to stay. And the more I look at it, the more it seems to me that I might be wrong. Looking at the data (Sentence 1), doesn’t it already say that more availability was based on more frequency?

I will investigate correlation and causality tomorrow.

Bis bald!