The path to 100% Using probabilities the right way
A few years ago, I wrote the following article, Probabilities that aren't
In it, I discussed the corrupting effect of the probability guesses that we all work with every day. I was reminded of it when I read David Spiegelhalter’s Does probability exist? Probably not — but it is useful to act as if it does. (Nature | Vol 636 | 19/26 December 2024) - in it, Spiegelhalter argues for one of my principles of asymmetric learning - obliquity - the intentionally unintentional consequences of process. That is, if we ask the right questions, in the right way, we will create more interesting observations, but they may not be the ones we expected.
Here’s my original piece:
The Probability of Technical Success, I am going to suggest, is a distracting, dangerous methodology.
Recommended by LinkedIn
(There may well be other reasons you can tell me, but I haven't yet heard a defence of the idea of a no-feedback-loop system.) Most conversations suggest that there is an awful lot of gaming, or body English, used to make sure the numbers that we get are the numbers that are wanted: 'you don't need a weatherman to know which way the wind blows...'
This may well seem a specious argument. 'Well, with what would you replace them? McKinsey designed our process 10 years ago, and we know it is cr**, but just imagine the effort to uproot it and start again...' Well, as they say, 'your lack of imagination is not an argument.' An industry that believes (or says it believes) in precision, in diagnostics, and statistical analysis, should be in a healthier place.
In Spiegelhalter’s piece, he writes:
The latter has something in common with frequentist definition of objective probability, just with the class of repeated similar observations replaced by a class of repeated similar subjective judgements. In this view, if the probability of rain is judged to be 70%, this places it in the set of occasions in which the forecaster assigns a 70% probability. The event itself is expected to occur in 70% of such occasions.
Key to our processes is to understand what we are doing the calculations for, and to rethink our processes. Belief in any number would bring the awkward realisation of how rarely our teams all agree on the accuracy of any of the numbers they’ve produced. The solution, as we argue with our path to market approach, is that we do not rely on any one composite, but create multiple potential paths: they may well each contain their own assumptions, but we’re establishing the process in order to create a path to learn, rather than to confirm or deny.
Manager at Biogen | R&D Portfolio Decision Analytics
1wPrecisely! And not just development paths, but risk paths. A conceptual mistake that I see around risk is that it always operates on some smooth mathematical landscape, but in our industry in particular, it is full of ripples. Risk discharge happens in chunks as uncertainties around candidate molecules are resolved. Because of that, there is a discontinuous set of discrete risk profiles which may reflect reality. A major departure from this reality is that pTRS adjustments tend to be done by treating it as a continuous variable, ignoring the steps/chunks that lead to discrete outcomes. For decision purposes, crafting the potential risk pathways with appropriate resolution will allow far richer insights than generating broad composites.
This is great! PoS for binary outcomes is a fun topic... Thanks for stoking this conversation and shining a light on a flawed process. ... surprising how much $ is swayed by a single # in a spreadsheet.