Everyday Research Methods

Could an Instagram filter predict depression?

One of the Instagram features the researchers coded was the filter people used on their photos. Photo: Travis Carr

This NBCNews online video explains a study whose claim is that "Your Instagram feed could predict if you'll suffer from depression." 

This headline presents an association claim, because "predict" is an association claim verb. The headline is not claiming that your Instagram feed could make you depressed (or not).  However, it is claiming that if we know something about your Instagram feed, we can predict whether you have symptoms of depression or not.

Take 2 minutes to watch the video now. As you watch, ask yourself, "what Instagram variable(s) predicted depression?"

Questions

a) What are the two variables in this claim? State them at the conceptual level.

b) How was each variable operationalized? Explain as carefully as you can. For each of these operationalizations, is it a measured variable or a manipulated variable? 

c) What kind of study do we need to support an association claim. Did the researchers conduct that kind of study? 

d) Sketch a graph or graphs of the study's results to the best of your ability. Will you sketch a bar graph or a scatterplot? Explain your choice. (Hint: The story mentions three variables, but the graphs you make will require you to focus on two variables at a time).

e) Can this study support the causal claim that "using Inkwell filters causes people to become depressed?" Why or why not? (apply the three causal criteria). 

f) Has this study been published yet? 

 

Suggested Answers

a) One variable is "way of using Instagram" and the other variable is "Level of depression."

b) The variable "Way of using Instagram" was operationalized via noting what kind of filter people used. There are several categories of filters, and it's not clear if the study categorized each filter separately, or if they quantified how "dark" or "color-free" the filter choice was for each photo. This was a measured variable.

Another variable the story mentions is the number of people in each person's Instagram photos. We can assume this was operationalized simply by counting the number of people in each photo. This is also a measured variable. 

The "level of depression" variable was also measured; the video did not indicate how depression was operationalized. We can assume that depression was measured via a self-report survey such as the Beck Depression Inventory, but you'd have to check the original article to be sure. Around minute 0:32, the scientist in the interview implies that the participants were actually diagnosed with depression, so it's also possible that each person had an interview with a clinical psychologist. 

c) A correlational study can support an association claim. This study was correlational because both of the variables were measured. 

d) You might have made a scatterplot with "level of depression" on one axis and "number of people in the photo" on the other axis, with a scatterplot sloping down from left to right (a negative relationship).

If you assumed that the "filter" variable was categorical, you might have made a scatterplot with "level of depression" on the y axis and bars labeled "Valencia" and "Inkwell" on the x-axis.  The bar for "Inkwell" should be associated with higher levels of depression. 

e) This study does have covariance, such that people with higher levels of depression were more likely to use black and white filters. It's not clear if the study has temporal precedence, because we don't know if the filters and the depression score were collected at the same time or not. This study also does not meet internal validity, because correlational studies cannot control for all possible third variables. We might wonder if some outside variable, such as gender, is associated with both the filters people use (perhaps women use more black and white filters just because they try to be more artistic) and gender is also associated with levels of depression (women tend to have higher depression averages than men, overall). We'd want to see if the researchers controlled for such demographic variables in their statistical analyses.  

Even though this study cannot support a causal claim, the finding is still interesting because it presents a possible behavioral predictor of depression. 

 f) The answer to this question is available at minute 0:20. 

 

 

Dogs in the MRI

Dogs were trained to lie still in this scanner long enough to participate in the factorial design. Credit: REUTERS/Bernadett Szabo

What's not to love about dog research? One recent study, conducted in Hungary and covered here by CNN, trained 13 dogs to lie perfectly still in a functional MRI (Magnetic Resonance Imaging) machine (this story describes how dogs learned to do it). 

After being trained for several weeks, the canine participants donned doggy headphones and heard different words pronounced in different tones of voice. Here's how the CNN journalist described the study: 

The study involved 13 dogs -- golden retrievers, border collies, a German shepherd and a Chinese crested -- who were trained to lie down and remain still for more than seven minutes while in a brain-scanning MRI machine.

 

The researchers used the machine to record and measure neural activity in the dogs' brains while they listened to a woman trainer, whom they were familiar with, recite various words in various intonations.

 

For instance, positive or meaningful words such as "well done," "good boy" and "clever" were said in both a praising intonation and a neutral intonation. Neutral or meaningless words, such as "even if," "although" and "however," also were said in both intonations.

...

Only when a praise word was spoken to the dogs in a praising tone of voice did the brain's reward center light up like a Christmas tree in the brain scans. The reward center is the part of the brain that responds to pleasurable stimuli, like food treats or being petted.

 

a) This is a factorial design. What are the two independent variables (IVs)? How many levels of each? How would you state the design (in ___ X ___ format?)

b) For each IV, decide whether it was within subjects or independent groups. Then use this information to identify the type of factorial this is (is it an independent groups factorial? A within-groups factorial? Or a mixed factorial design?)

c) What was the dependent variable, or DV, mentioned in the excerpt above? 

d) Sketch a graph of the results that the journalist described. (Hint: Put the DV on the y-axis, as usual. Pick one IV for the x-axis, and another IV for different-colored dots or bars.) 

e) Inspecting the graph you created, is there an interaction in these data? What about each main effect--would you predict they would be significant?