Hey, sorry for the late post, had to give a talk this morning!
I’ve written a bit about QALYs here before, or Quality-Adjusted Life Years.
Basically, hospitals need to make decisions about how to use limited resources. They’d like to run some sort of cost-benefit analysis that can collapse across all the relevant dimensions of health—lives saved, diseases cured, disabilities “““normalized,””” (is that enough scare quotes?) and so on, into one clean measurement.
QALYs have emerged as a sort of blueprint for what this sort of metric would look like, and of course there are a bunch of minor variants, but the basic idea is pretty straightforward. You hand out surveys to a bunch of folks to see how many years of, say, blindness they would consider equal to 10 years of “perfect health.”
In practice, the average answer you get is roughly this:
20 years blind = 10 years of perfect health.
Great! So now we apply an adjustment: Each year lived by a blind person is worth half as much as a year lived by a person in perfect health.
Mathematically, this means blind folks are basically half-dead, because “upgrading” them from dead (0) to blind (0.5) is the same as ““upgrading”” them from blind (0.5) to “““perfectly””” healthy (1.0)
(Sorry for the national scare quote shortage everyone, it’s just a supply chain issue.)
We’re just multiplying, so the math here is nice and tidy.
If you save a blind person’s life and they get an extra 20 years, that’s 0.5*20 = 10 QALYs added. And if you restore a blind person’s sight for 20 years, that’s also (1-0.5)*20 = 10 QALYs.
So now we can just ask, “Which intervention is cheaper?” and prioritize that one. Ta-da! Cost savings! Evidence-based decision-making!
I recently used this as an example of how the metrics we use make tradeoffs for us.
Say that after some corporate “restructuring” we record a 10% annual increase in QALYs added. Yay, we did it!
But what did we do, exactly? Here’s an uncomfortable way of drawing out the problem:
Did we make people healthier?
Or did we just do more eugenics?
Maybe a little of Column A and Column B?
Who knows! The numbers can’t help, because QALYs bury this distinction between improving people’s health and leaving “unhealthy” people behind.
And the thing is, blind people think their own lives are going pretty damn well, thank you very much! They don’t think they’re half-dead, and they’re a lot happier than you might imagine. So the survey results from everyone are probably recording ableism more than anything.
And that’s what makes it so weird that when I recently presented this example to a philosopher, their immediate response was, ”Well at least now your decisions aren’t biased.”
I think they were going, well now it isn’t just up to individuals how much they arbitrarily value, say, restoring sight vs. saving blind folks’ lives.
But uh…you know systems can be biased too, right?
QALYs are absolutely a deeply biased measurement, in countless ways.
And because QALYs standardize biases, what we’ve removed is variance!
If you’re blind, you will always be treated as half as valuable. Period.
Is this an improvement? It’s not so obvious!
If I’m blind, maybe I’d rather take my chances trying to find a less ableist doctor who has more familiarity and experience and compassion for the blind community.
This is probably the single biggest takeaway I’ve learned in my first quarter at Hopkins: The strength of standardized measures and procedures is their consistency. And of course, that’s their weakness too, when consistency becomes rigidity!
That’s why any metric (or even suite of metrics) guarantees some sort of value capture. Our values aren’t fixed targets out there just waiting for the perfect measurement system to come along and map onto them.
But the larger point is that standardized procedures also guarantee value capture in essentially the same way. Consistency makes you rigid, and suddenly your processes fetishize their own completion on paper over the reality out there. (Witness the endless proceduralism I ran into trying to get on a health care plan!)
That’s why I’m sitting in on a couple graduate seminars in Anthropology this semester. If I wanna get beyond reality on paper, to get past the artificially cleaned-up versions of AI research I’ve been getting from folks at conferences and in coffee shops over the last couple months, I need to enter these spaces and see what people are actually doing day-to-day. And I need the tools to study their ordinary practices carefully, without simply falling headlong into their familiar procedures.
Here are the two main slogans I’ve learned so far:
Words are not Concepts are not Things
Half of Participant Observation is “Participant”
So that’s a taste of what you have to look forward to, and yes it absolutely will culminate in a bizarro anthropology of how my cousins evaluate the most scalable offensive player we’ve ever seen, Luka Doncic, Mr. 540 dribbles per game, and why I think we just mean different things when we ask who the “best” player is. (But my definition is better.)
So uh, stay tuned for that! Cheers.
Commentaires