What Does “Predict” Mean in Research?

In her TED talk, Amy Cuddy says,

Nalini Ambady, a researcher at Tufts University, shows that when people watch 30-second soundless clips of real physician-patient interactions, their judgments of the physician’s niceness predict whether or not that physician will be sued.

This is quoted all over the place, yet I have been unable to track down the study. Maybe it is unpublished or in press, or maybe it is published under a title that doesn’t mention physicians or videos.

In the meantime, I wonder whether Cuddy might have conflated two separate studies by Ambady: one of surgeons’ tone of voice (2002), and another of soundless clips of teachers (1993).

But my greater concern is with the word “predict.” As Cuddy puts it, the “judgments of a physicians’ niceness” actually predict whether or not that same physician will be sued in the future. To determine this, a researcher would have to follow the physicians over the long term and compare their subsequent lawsuit patterns (if any) to the initial ratings.

That would be terrifically difficult to accomplish. First, doctors with malpractice litigation histories are a small percentage of the whole, so the sample size would be tiny. Second, how long do you wait for a doctor to be sued? Two years? Five? Ten? Indefinitely?

Instead, I suspect the study followed a procedure similar to that of “Surgeons’ Tone of Voice.” (If I find out I am wrong, I will post a correction.) That is, the ratings of the videos were related to doctors’ existing lawsuit history. If there was any prediction, it was retrospective; as the authors state in the surgeon study, “Controlling for content, ratings of higher dominance and lower concern/anxiety in their voice tones significantly identified surgeons with previous claims compared with those who had no claims” (emphasis mine).

Why does this matter? There’s a big difference between predicting past and future events. People are easily dazzled by the idea that a thirty-second video clip (or sound clip, or whatever it may be) can predict a future lawsuit. The idea that it might predict a person’s existing history is perhaps interesting (if it holds up under scrutiny) but less dazzling.

When it comes to doctors, I imagine those with a lawsuit history might be a little grumpier than the others. I can also see how tone of voice could affect patients’ sense of trust and comfort. But I know of no study that demonstrates that ratings of soundless video clips of physicians predict whether they will one day be sued.

To avoid turning scientific research into a magic show, use the word “predict” carefully and precisely. Also, give a little more detail when referring to research, so that those interested can look up the study in question.

Update: I just learned that Ambady died in 2013. See the comments below. Thanks to Shravan Vasishth for the information.

Another update: Thanks to Martha Smith for explaining various  kinds of “prediction” in statistics. (See here and here.)

Next Post
Leave a comment

3 Comments

  1. Why don’t you ask Nalini?

    Reply
  2. Oh, sorry. It seems she is no more. She died in 2013.

    Reply

Leave a Reply

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

%d bloggers like this: