(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
. . . . . . . . . .



The Marshmallow Test is BS: A Review of Escape From Model Land [1]

['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.', 'Backgroundurl Avatar_Large', 'Nickname', 'Joined', 'Created_At', 'Story Count', 'N_Stories', 'Comment Count', 'N_Comments', 'Popular Tags']

Date: 2023-05-05

The marshmallow test is some bullshit.

First, it is classicist and therefor racist and thus inaccurate. But no one should have taken it seriously to begin with. Any child with half a minute of lived experience knows that adults lie. The chances that they are going to get the second marshmallow are pretty much nil in the average kid's experience. Mom is going to take them home early, or some worthless goody-two shoes adult is going to give them broccoli or some other crap so they can be healthy, or its gonna be some lame joke like the time their uncle conned them out of ten bucks by breaking the twenty they got for their birthday (still not sure how that happened, but you better believe their picking out old folks homes for Uncle Mike in their heads. Unfunny jerk is gonna get his. Yes he is.) But scientists and economists couldn't see this simple truth because they lived in Model Land and thus were astonished when kids didn't do what the model said they should.

The above example, slightly embellished (It's not cool stealing from six-year-olds, Uncle Mike), is perhaps the best distillation of the argument in Escape from Model Land by Erica Thompson. Thompson's argument is straightforward: models are not as helpful as they are made out to be. But they could be made more useful if we applied some basic rules to their use.

Using financial, covid-19 predilections, weather forecasts, and climate models, Thompson walks through the various issues that come with living win what she calls Model Land -- putting too much faith in models or in decisions made by models. Thompson demonstrates points out that a model is not reality, and thus has limitations that require decisions to be made about how to deal with those limitations. And those decisions affect both the model outcomes and the perception of fairness and accuracy surrounding the models. And because the models do have biases built into them, and overreliance on models can be detrimental to public trust in expertise and damaging to democracy.

Thompson has several solutions: be clear about the decisions made in the models. Be clear about what the model is actually supposed to help decide and why it is appropriate for that decision. Use a lot of models form many different perspectives so as to engender more trust in the process from all involved. All of which is good advice, but Thompson doesn't address a significant driver of distrust in models and science: lies.

One of the reasons that faith in expertise has dropped is because experts have failed, or been seen to fail. Thompson discusses the models that lead to the British government deciding that Iraq had weapons of mass destruction in terms of the failure of the intelligence services. But she doesn't bring up the lies at the heart of the data fed into those models. She also doesn't discuss the fact that the right wing in the anglosphere has been part of a decades long project to close itself off from sources of truth that conflict with its ideological priors. Having models from all perspectives cannot help restore trust in expertise when an entire well-funded movement exists to tell its adherents that there is no such thing as expertise.

Despite this flaw (and to be fair, there may not be anything modelers and people who depend on models can do about the situation), the book overall is worth buying. Thompson makes a clear case for why Model Land is a dangerous place, but also for why it is sometimes beneficial to visit and how to make it less dangerous. Anyone who is interested in how models shape our world, which should be any engaged person, and how we can avoid the associated pitfalls, should pick up a copy.

[END]
---
[1] Url: https://www.dailykos.com/stories/2023/5/5/2167573/-The-Marshmallow-Test-is-BS-A-Review-of-Escape-From-Model-Land

Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/