A Walk in the Forest to Discover the Bayesian Universe !
The Turing Cursor
The Turing Cursor
This article introduces a tool that we used in a previously presented educational activity.
In a new game mode, a cursor appears when the player approaches a mushroom. It starts at an initial position corresponding to 0. This represents the prior belief we have before conducting the first test on a given species.
There is no reason to assume whether the mushroom is good or bad before testing it.
If the test result is positive, the cursor shifts one position to the right, indicating the probability of encountering a mushroom species that produces a treatment.
On the other hand, if the result is negative, the cursor moves to the left, implying a higher probability of it being a bad type of mushroom. And that's precisely the principle of Bayesian probabilities (or inference) !
The idea is to start with a prior belief, which can be different from 0 in general, and each experiment updates our prior belief. The measurements we make will successively move the cursor to the right or left.
Since the statistical aspect of the presence of the beneficial molecule was highlighted in the first part, the cursor may change direction, sometimes dramatically!
The student/ player now has a tool derived from the scientific method to maximize their score. Initially, the cursor helps keep track and remember whether there have been more "+" signs than "-" signs. This tool is extremely valuable and already enables making more informed choices.
In the next part of the game, the player will have a new tool at their disposal. We have added a blue bar to the cursor, representing the threshold beyond which the probability of finding a good mushroom family is sufficiently high to make a "bet" with a positive expected gain, taking into account the ratio of points lost and points gained in case of failure or success.
In other words, this bar allows for measuring risks and making the most informed choice based on the accumulated data. There will never be certainty, of course, but we now have a tool that provides a reliable estimation.