Lightnews — Scholar-powered news

Deena Mousa @deenamousa.com · 1d

Read the full post here: newsletter.deenamousa.com/p/when-more...

When more automation means more human workers

A look at elasticity of demand

newsletter.deenamousa.com

Deena Mousa @deenamousa.com · 1d

There are exceptions to this rule, though; the rise in mobile banking eventually did reduce bank teller employment. If the technology fully replaces human inputs, or drives efficiency gains so large they outpace the increase in demand, jobs will be lost.

1

Deena Mousa @deenamousa.com · 1d

In each of these cases, employment holds steady, but the skill mix changes. The work gets more complex as machines take over the simplest tasks.

1

Deena Mousa @deenamousa.com · 1d

This can happen in other fields too — for example, spreadsheets didn't get rid of accountants, but rather made them more in demand than before. In law, predictive coding and e-discovery tools reduce grunt work but let firms take on more complex cases.

1

Deena Mousa @deenamousa.com · 1d

This is an example of Jevons paradox: when tech makes something cheaper or more efficient to use, we'd expect to use less of it and achieve the same result. But in some cases, people use *more* of the good because it's more accessible.

1

Deena Mousa @deenamousa.com · 1d

When ATMs spread across the U.S., you'd expect bank tellers to be laid off en masse. But teller employment actually rose during the 2000s, growing 2% annually.

By reducing the cost of operating bank branches, ATMs made it possible for banks to open more locations.

1

Deena Mousa @deenamousa.com · 1d

New post: When can more automation mean more human workers?

One argument I made in my recent @worksinprogress.bsky.social piece is that if automation made reading scans quicker and cheaper, this might result in *more* jobs for radiologists, rather than fewer.

How does this apply to other jobs? 🧵

1 2 3

Deena Mousa @deenamousa.com · 13d

If you enjoyed the article, I write about health, economic development, and scientific and technological progress here: newsletter.deenamousa.com/

Under Development | Deena Mousa | Substack

Writing about global health and development. Click to read Under Development, by Deena Mousa, a Substack publication.

newsletter.deenamousa.com

1 1 12

Deena Mousa @deenamousa.com · 13d

AI radiology today is powerful, but it consists of many narrow islands of automation that have failed to replace radiologists' time.

This isn't the full picture. Read more in the @worksinprogress.bsky.social piece: worksinprogress.co/issue/the-a...

The algorithm will see you now - Works in Progress Magazine

Radiology combines digital images, clear benchmarks, and repeatable tasks. But replacing humans with AI is harder than it seems.

worksinprogress.co

3 5 28

Deena Mousa @deenamousa.com · 13d

On top of this, datasets are still a barrier for some scans and patient populations. Even with 700+ FDA-approved imaging AIs, they cover only a fraction of real-world work; most cluster around a few conditions like stroke, breast cancer, or lung cancer.

1 25

Deena Mousa @deenamousa.com · 13d

There’s another catch: many models struggle outside the hospital they were trained in. In 2024, 38% of FDA-cleared radiology AIs were tested on data from just one hospital. Move them elsewhere, and accuracy can drop by up to 20 percentage points.

2 16 62

Deena Mousa @deenamousa.com · 13d

Even platforms that bundle multiple models still spit out a list of disconnected yes/no answers.

But a good radiologist integrates findings, context, and patient history into a single coherent view of a person's health.

1 2 33

Deena Mousa @deenamousa.com · 13d

That means a radiologist can’t just “hand off” a scan to AI. To cover a typical day, they’d need to pick from dozens of different models, run each one separately, and stitch the answers together.

1 1 28

Deena Mousa @deenamousa.com · 13d

One reason for this: radiology AI doesn't read and form a holistic view on a scan like a doctor does.

Each model is built to answer a single narrow question:
– Is there pneumonia in this chest X-ray?
– Is there a lung nodule in this CT?
– What’s the calcium score in this scan?

2 5 49

Deena Mousa @deenamousa.com · 13d

If benchmark performance translated to reality, radiologists should be the canary in the coal mine of AI job loss.

Instead, radiology residency slots hit a record high in 2025.

1 2 29

Deena Mousa @deenamousa.com · 13d

CheXNet, released in 2017, could detect pneumonia better than a panel of board-certified radiologists.

Since then, hundreds of models have outperformed humans on benchmarks. Today, over 700 FDA-cleared radiology AI models exist. They make up >3/4 of all medical AI devices.

1 28

Deena Mousa @deenamousa.com · 13d

In 2016 Geoffrey Hinton said “we should stop training radiologists now" since AI would soon be better at their jobs.

He was right: models have outperformed radiologists on benchmarks for ~a decade.

Yet radiology jobs are at record highs, with an average salary of $520k.

Why?

8 56 140

Deena Mousa @deenamousa.com · 21d

Full set of winners here: www.experimental-history.com/p/blog-extr...

Blog Extravaganza 2025: the winners

YEEHAW

www.experimental-history.com

1

Deena Mousa @deenamousa.com · 21d

And, self-servingly, I appreciated the honorable mention among such an interesting set of articles

newsletter.deenamousa.com/p/how-much-...

1

Deena Mousa @deenamousa.com · 21d

I'm also a big fan of intentional self-experimentation, so I really enjoyed this piece about identifying lactose intolerance as a driver of morning migraines.

substack.com/home/post/p...

1

Deena Mousa @deenamousa.com · 21d

I loved this piece by Rabbit Cavern about Cheetos' brief anarchist movement and am looking forward to future rabbit holes

rabbitcavern.substack.com/p/did-cheet...

Did Cheetos try to incite a rebellion in 2008?

A flamin’ hot tale of edible anarchy

rabbitcavern.substack.com

1

Deena Mousa @deenamousa.com · 21d

I really enjoy Adam Mastroianni's Experimental History — I expect I'll like following some of the winners of his blog competition just as much! A few of my favorites in thread:

1 1

Deena Mousa @deenamousa.com · 21d

And I'm a big fan of intentional self-experimentation, so I really enjoyed this piece about identifying

substack.com/home/post/p...

Deena Mousa @deenamousa.com · 21d

I loved this piece by Rabbit Cavern about Cheetos' brief anarchist movement and am looking forward to future rabbit holes

rabbitcavern.substack.com/p/did-cheet...

1

Deena Mousa @deenamousa.com · 22d

api.omarshehata.me/substack-pr...

Why do radiologists still exist?

AI's first white-collar casualty hasn't fallen yet

api.omarshehata.me

1