Minds Over Matter

by: Ryan Quan, Data Scientist @ Omada Health

Pop Quiz! You’re marooned on a desert island with no hope of rescue for several months. Which outlook sounds most like yours?

A) I’ve watched enough Bear Grylls to figure things out. I bet I could survive until the next boat arrives!

B) Doh! I’ve never fished with my bare hands or made a fire without matches. I’m probably a goner.

Most folks fall easily into one camp or the other — and these kinds of beliefs about ourselves are called mindsets.

Stanford University psychologist Carol Dweck, author of Mindset: The New Psychology of Success, proposed two different types of mindsets: fixed and growth-oriented.

People with fixed mindsets tend to avoid challenges. To them, failure suggests that they lack the “chops” required for the task. People with growth mindsets, however, tend to embrace challenges. They believe they can learn to conquer just about anything — even something as tough as behavior change. (Studies from little old places like Harvard have shown that your mindset can profoundly impact your life. Even your health.)

So do mindsets matter for weight loss? Let’s find out by trying to quantify them.

Mindsets in Prevent®

Here’s a hypothesis: Participants with growth mindsets will do better in the Prevent program. Sounds like interesting idea, but how can we even begin to distinguish between participants with fixed and growth mindsets? Sounds like a fuzzy question requiring a fuzzy answer. So let’s dig into the wonderful world of text data.

All of our Prevent participants are asked the following question in their entry questionnaire: What challenges do you expect to face as you try to change your eating and physical activity habits?

Here are the most frequent words in the responses, split by participants who met their weight loss goal and those who did not.

By including stop words like “am”, “have”, and “is” — meaning words that are commonly removed in natural language processing — we’re able to observe some interesting patterns around past and present tense.

Present tense (“this is”, “I am”) correlated with less weight loss.

Past tense (“I have been”, “I was”, “I used to”) correlated with more weight loss.

How on earth can grammatical tense relate to weight loss? Sounds like a lot of pop-science baloney, doesn’t it? But instead of just dismissing the notion, let’s dive into some example responses from our participants. Remember, participants are talking about their perceived challenges to modifying dietary and physical activity habits (red = no weight loss, blue = 5%+ weight loss):

Notice anything? As Dr. Dweck mentions in Mindset, people who are growth-oriented often describe traits and tendencies in the past tense, which “releases one from the tendency to expect the same outcome to happen again.” On the other hand, people with fixed mindsets tend to describe traits and tendencies in the present tense as a way of professing their identity. “This is how I have been so far” gives a lot more room for improvement than “This is who I am now”.

Quantifying Mindsets

So is there any truth to the power of mindsets in helping participants achieve their weight loss goals? Do people who hit their weight loss goals tend to frame their negative tendencies in the past?

You tell me.

The red line represents one Prevent participant who did not see much weight loss, while the blue line represents a second participant who did. Individual results, of course, vary from participant to participant.

Comparing the two lines above, we see some trends worth mentioning. The participant who lost over 10% of his initial weight tended to talk more about the past. Is this a sign of a growth mindset? Someone who sees their challenges in the rear view mirror, and looks ahead to positive outcomes? The participant who didn’t lose weight, however, saw much greater variation, and far fewer messages written in past tense, on average. Are they stuck with their present tense conceptions of themselves? Is their mindset in the midst of changing somehow?

Perhaps. This is, of course, just one example of the type of information we look at here on the the data science team at Omada.

As the number of participants and data points grows along with our desire to understand what matters and what doesn’t for health behavior change, we plan to leave no data untouched. Nope. Not even grammatical tense.

Have any thoughts you’d like to add? Highlight some text and leave us a note! We’d love to hear from you.