Computers can't replace humans, in education or in breast cancer detection: a tale of two studies

First up, breast cancer:

The companies that make the nearly $175,000 machines, the doctors who felt obligated to use the machines, and the women whose mammograms were interpreted by the computers should not be surprised that the CADs (computer-assisted detection) failed to screen any better than the human physicians. Instead, just as I’ve been writing and others have been saying, 20% more women were subjected to heightened anxiety over the possibility that they had breast cancer because the CADs indicated that they needed invasive follow-up biopsies, which, in only one out of 2000 cases, turned out to be a cancer. (How many of those women do you suppose were also offered anti-depressants and all those other drugs you see advertised to make women smile again? Yes, that’s snarky of me but it’s exactly what I think about when I hear about these studies.)

Read the New York Times article here (it’s good). And read the abstract of the study, appearing in the New England Journal of Medicine, here.

Next, education:

Elementary schoolchildren who utilized educational software in the classroom for math and reading didn’t demonstrate any greater mastery or deficit, according to a study just released by the US Department of Education’s National Center for Education Evaluation and Regional Assistance (NCEE) (one of the four centers of the Institute of Education Sciences, which, by the way, for you smaller government-loving conservatives, was created by President Bush in 2002; it’s budget request for FY 2008 is $594.3 million, or, as Matt Dole prefers, $594,300,000 for a Bush-created, Bush-maintained entity, out of a total USDOE budget request of $56 billion – a decrease from 2007, of course, because states just love unfunded education mandates; the study we’re reading about today probably falls within the $282 million for Statistics and for Research, Development and Dissemination).

Now, I’m an education and statistics and research lover, so none of this bothers me as a general proposition. What I really need to know and analyze, however, is which if any Bush favorites get and got the money for providing the software – that’s what intrigues me.

Read more about the study in this Seattle Times article and read the report here.

The part I love the most: No Child Left Behind mandated that the study take place, but none of the news articles indicate how the study was financed, how much money it received and who got the money (let alone provide any of the info or links I provided above).

After glancing through the report to Congress made by the governmental entity charged with the study, I’ve sent in a request for financial figures related to the cost of the study (because I’ll allow that the info might be in there but I couldn’t find it).

I’m not questioning the value of the government entity that guides and reports on this and other studies. From a ten minute browse of the website, the people and the place seem credentialed and serious. Anyone with more info, please contribute.

However, I have to put these two studies into the growing pile of studies I can recall that tell us things we alread know or could have expected to be the case: just like the breast cancer detection via computers study, when it comes to humans, humans do it better.

When it comes to products, I’ll allow that computers may have the John Henry vs. the steam engine edge. But when we’re talking people, people do people better.

9 thoughts on “Computers can't replace humans, in education or in breast cancer detection: a tale of two studies

  1. Breast CancerCommon Breast Cancer MythsThe first myth pertaining to this disease is that it only affects women.Second myth that is associated with this disease is that if one has found a lump during an examination, it is cancer.Third is that it is solely hereditaryThe next myth associated with breast cancer is downright ridiculous. Would you believe, that in this day and age, some individuals still think that breast cancer is contagious?Conversely, some individuals foolishly believe that breast size determines whether or not one gets cancer.Finally, another myth that is associated with this disease is that it only affects older people. This is not so. Although the chance of getting breast cancer increases with age, women as young as 18 have been diagnosed with the disease.You can find a number of helpful informative articles on Breast Cancer at breast-cancer1.comBreast Cancer

  2. Comparably, but not irreplaceably, when it comes to certain tasks. And hey, how many tasks could each of us come up with that we WISH computers would do instead of us? Thanks.

  3. But I think you get my point, no? Yes, I get your point. (And, come to think of it, I might have been a little off-topic.) However, I just didn’t want to exclude the possibility that machines might some day perform comparably to humans in breast image analysis.

  4. Anon – no, I didn’t know that specific info, and I’m not surprised, and I know it’s important to recognize where innovations like that fit in. But I think you get my point, no?

  5. That’s really interesting, Paul. I confess, you don’t have me convinced about the computer-assisted stuff. Especially since I just heard the doc confirm for me what I knew: MRIs for breast detection, right now, should not be used for all women. It’s not a matter of a false positive – it simply can’t tell you what it sees. The docs tell you.I’m sure though, as you imply, some combo of all these tools will be better perfected. But I’d also love for us to figure out some prevention methods too.

  6. Westender – that last paragraph – that’s exactly what I was going for. There’s intuition that develops after experience and you can’t replace that with a computer. That’s all I’m saying – I’m a big chaos theory person so I just don’t trust any one system for anything.

  7. But when we’re talking people, people do people better.You do know that the vast majority of complete blood counts (in which the images of the various types of blood cells are identified) are done by machines, right? That’s why a CBC costs only $2. (In special cases, a medical technologist will do a manual CBC.)

  8. I’ve been a computer geek since the 1960s, studied computer science in college, and made a decent living deeply imbedded in the technology world. I have faith that computer assisted analysis of body imaging will continue to improve and eventually become much more reliable than human eyeballs, even when the eyeballs are attached to a skilled radiologist. It is common for a particular technology to start off being much less efficient than even a marginally adept human, yet over time be improved to the point where the technology performs its limited task at a level beyond all but the most skilled and gifted humans. I am quite confident that there will be a time when no physician would consider his/her examination to be complete without having a computer assisted assessment.In the case of disease detection, a false positive is a much better error state than a false negative. I understand your point about the seemingly needless anguish of a false positive, but we would rather have the diagnostician err on the side of conservatism every time. It is not unusual for a human diagnostician to make this kind of mistake.The mistake that kills is the false negative. If machines can be improved to the point where they help detect cancers while the tumor consists of tens or hundreds of cells, enabling much earlier treatment, wouldn’t a few false positives be worth it?A friend of mine, Professor Marlin Thurston is the inventor of the Neoprobe technology, a device used during surgery to help the surgeon find cancer cells that have been missed on the imaging studies, and would probably be missed by the doctor during surgery. Dr. Thurston came up with machine when was himself a cancer patient, and gained an understanding of the difficulty oncologists and surgeons had in find all the cancerous cells on the operating table. A particularly nasty kind of false negative is the one that happens when you go through the physical and emotional trauma of cancer surgery without getting it all because it couldn’t be detected by human senses.You make one point with which I wholeheartedly agree – when the government doles out money for research in technology, or research in anything else for that matter, how do we gain confidence that the money went to the most qualified investigators, or that the research merits getting any government funding at all? Sometimes really good stuff happens as the result of government-sponsored research, like walks on the moon. But all kinds of money is also spent on stupid stuff just because it repays political favors.And it will keep on happening until the people rise from their apathy and demand better for our country.PL

  9. But is it really human vs. machine if machines are tools which are made by and used by humans? Perhaps the issue is not whether a machine is used, but how disconnected the human user is from the machine.In other words, if a physician puts her mind on standby and just relies on a machine, she’s essentially taking the human out of the equation. But a physician who understands that the machine is a tool to enhance human abilities is still keeping herself in the equation.I was talking to a physician once who remarked that Indian clinicians typically have superior clinical skills and she surmised this was the case because physicians trained in India don’t have access to high-falutin’ machines and high-tech equipment, so they can’t order tests for everything under the sun. Instead they have to rely on their skills and judgment more than western physicians.

Comment here

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s