Awarded NORC Pilot Grant to Develop Mechanical Turk Capacity at UAB

Consider a rodent study that uses a high fat diet to induce obesity. What do the results represent? Do they model the effects of all high fat diets in humans? Do the results provide insights into the effects of obesity? Or should we limit the scope of inference to consider only that specific diet within rodents?

The use of language in communicating research results can influence the way others interpret them. Analyzing the use of language in research reporting can be a time-consuming task, though, which can be difficult to automate because it often requires human judgment. The Nutrition Obesity Research Center at the University of Alabama at Birmingham has recently awarded me a $25,000 Early Career Study Award to explore using Amazon’s Mechanical Turk to crowdsource the evaluation of language used in scientific publications.

In particular, I will be looking at the use of language in studies investigating high fat diet induced obesity in rodents. These models entail using a high fat diet, often mostly composed of lard, as a model for obesity. Although these models can be useful, and I have personally been involved in projects that use them, the model comes with important caveats. Not all dietary lipids are the same, meaning a diet containing other lipid sources may result in different metabolic manifestations. Similarly, diet is not the only means of developing obesity. Using Mechanical Turk, I aim to quantify how often research using high fat diet induced obesity models make sweeping generalizations beyond the model limitations.

This research will serve two purposes. First, it will be a practical means by which to build capacity in using Mechanical Turk to evaluate research. Preliminary results using Mechanical Turk that looked at human nutrition and obesity research shows that the method is promising, but my preliminary work had not been nearly as in depth as the present study. Second, evaluating the misleading use of language in animal models of obesity may highlight an important avenue through which to improve nutrition obesity research to better communicate what we collectively know and don’t know. Alternatively, the results may demonstrate that researchers in fact do communicate results with appropriate caveats, which would indicate that common confusion over animal models of obesity may be derived elsewhere.