“I’m shocked, shocked to find that cheating is going on in here!”

A couple of weeks ago I was shocked —shocked, I tell you — to find a headline in the Boston Globe reading “Harvard, MIT researchers find cheating in online courses.” Imagine that! You have an online course, where students are unproctored, and we’re surprised to hear that some of them cheated? How naive can we be? This has long been a concern of mine about testing outside of the classroom, whether it’s in an online course, a take-home test, or a project.

Identifying the problem is easy. Identifying the solution is much harder. In fact, I don’t have a solution, other than to recommend a healthy amount of skepticism when assessing students’ credentials that are based on any unproctored situation.

Here are a few clips from the Globe’s article:

Certain users were answering test questions “faster than humanly possible.” Turns out, the users were cheating. 

…hundreds of learners were skirting the system by using multiple accounts to cheat on tests for massive open online courses, or MOOCs. Users would take a test with one account and find the wrong answers, the scientists said, then use another account to take it again and plug in the right answers, “lickety-split.”

Harvard and MIT launched the edX startup three years ago, and the free, online courses are wildly popular. Users from around the world can now study everything from medieval liturgy to computer science, and earn certificates.

MOOCs currently enroll 5 million people, and other universities have joined edX and other similar ventures.

They found that 1.3 percent of certificates earned — a total of 1,237 certificates — appear to have been obtained through that type of cheating, a percentage they said is small, but significant, and a pattern Ho called “wholesale falsification of a certificate.” 

Just as MOOCs are a first-of-their-kind type of learning, this type of “evil twin” cheating is a first, they said, and professors are still trying to understand its implications and how to combat it.

The paper includes suggestions such as randomizing questions and not releasing answers until after assignments are due. Researchers said they want to make sure any anticheating strategy doesn’t compromise learning.

“We seek to reduce the potential for cheating, but we also try to maximize the opportunities for learning,” Ho said.

Cheating was more prevalent among users who earned 20 or more certificates, they found, where 25 percent of users appeared to have cheated. It was also more prevalent on government, health, and social science courses, rather than in science, technology, and math classes, the researchers found.

Cheating was highest among young, less-educated males outside the United States, they found, something they said could correlate with the perceived value of the certificate in different countries.

Categories: Teaching & Learning