A free, online test for cognitive impairment

A friend told me that he had taken some online tests for depression and cognitive impairment.  He is almost as old as I am and was curious about what the tests would find.  They found that he was not depressed, and he was not cognitively impaired.  Had they asked me, I'd have told them the same.  The test uses AI to analyze your discussion of a picture to find markers for cognitive impairment.  I know there has been a lot of research in this area.  In fact, I've participated in some.

The idea was intriguing, so I thought I'd give it a try too.  It's provided by a company called LANGaware, and no referral or medical review is required.  They claim, "LANGaware is a clinically validated cognitive health screening tool powered by AI and patented machine learning.  Analyzing language and speech patterns, LANGaware offers insights into your brain health, delivering probability results for cognitive health conditions."

You take the short test, the AI robot interprets it, and you get the results in a few minutes.  If you're disturbed by the results, it's up to you to take it to your medical provider.  What could be easier?

I am tested by researchers at the University of Washington's Harborview Medical Center every week, and I feel well about how I do on those tests.  So I wondered what this AI robot would find.  

Well, it said I am in the middle of the cognitively impaired range.

In taking the test, the first thing I did was inadvertently take a test on depression.  Not surprisingly, it found that I'm not depressed, but ... was part of the test to see if I could follow instructions to get to the right test?

The cognitive test was actually familiar to me, although I never knew how I was evaluated.  Since 2015, I've had comprehensive cognitive testing every year or two, although we are currently skipping two years before my next test.  In this test, I was shown a picture and was told to explain verbally what I saw going on in the picture.  The picture was familiar, and I could recall what I'd seen in the same picture different times over the years.

So I was surprised when I found that I hadn't done well.  There was a discussion of eight different parameters, and I figured I was evaluated using them.  When I read about each one, I concluded that I hadn't spoken formally enough, not using complete sentences with good grammar, etc., and that's why I had scored poorly.  You're allowed to take the tests as often as you like, so I took it again.  This time I was careful to use my very best diction and grammar.  (In Beating the Dementia Monster, we discuss how the practice effect can confound cognitive testing.)

The second time, I scored in the normal (not impaired) range, but I noticed that the results evaluation showed me different parameters used for evaluation.  I'm guessing I was actually evaluated on even more parameters, but, for some reason, these were the most interesting for this taking of the test.  So while my overall score put me in the "normal" range for cognition, I was graded down on "lexical variation," "reference pointers," and "duration of speech."  Only "reference pointers" was mentioned in the first test.  When I was graded down, a score was marked as, "a participant with significant cognitive decline."  Ouch.

On the other hand, the second AI robot may have taken a different approach to evaluating me.  This might be why it got a different result.  But I don't know.

Here's a graphical representation of my first test:

If you can see it, the largest colored area is healthy people.  The purple triangle in the middle represents some cohort of patients with cognitive impairment.  I'm the green; more OK than the patients with cognitive impairment, but a long way from the healthy cohort.

I'm thinking I'll take the test again tomorrow.  No telling what I'll find.