McLaren Receives NSF Grant to Data Mine Learning from Erroneous Examples

June 19, 2017

Erroneous examples, step-by-step examples of incorrect problem solving, is a pedagogical approach used in only a few fields, such as medical education. Bruce McLaren, an associate research professor in the Human-Computer Interaction Institute, recently received a grant of just under $1 million from the National Science Foundation (NSF) to research why learning from erroneous examples is successful and how it might be integrated into instruction more generally.  

McLaren and his collaborators will use educational data mining techniques to uncover and analyze student actions as they interact with erroneous examples.

"Learning through erroneous examples is a novel, little-used instructional technique," said McLaren. "The approach is in some ways counterintuitive and counter to Skinner psychology, which posits that students learn by mimicking correct actions. However, some researchers believe that challenging students with errors, prompting them to reflect on and correct errors, has significant instructional advantages." 

As a learning science researcher, McLaren is interested in how technology, specifically educational software, can support education. His past projects have investigated other educational technology questions such as: Can educational games lead to learning in mathematics? (Hint: yes, they can.)

McLaren’s new project will apply educational data mining techniques to log data collected from over 2,000 middle school mathematics students studying decimals. McLaren and his colleagues will employ data mining techniques to identify patterns in the data sets at a micro level.

The focus on data mining is part of a trend in the learning sciences to leverage data as a means of understanding how people learn and how to make instructional material more effective.

The research team will use an iterative plan, analyzing the log data of online learning from erroneous examples, then revising and repeating to explore whether enhancements to the erroneous examples improve learning. 

"We want to improve the materials over time," explained McLaren. "Due to the complexity of revising materials, in particular, the many different ways the materials can be modified and enhanced, it is necessary to work deliberately and iteratively to find ways to improve the instruction."

He continued, “I’m really excited about this project because it’s an opportunity to dig deep into data, to explore the underlying reasons for how and why students learn with erroneous examples. So often in research we get interesting results from experiments, but do not fully understand the results. Here, we have an opportunity to find out how and why students benefit (or do not benefit) from our instructional materials.”

McLaren's NSF research grant will contribute to the variety of learning science research that happens within the Human-Computer Interaction Institute. You can also learn more about the  human-centered data science tract and learning science master's programs.