Predicting A Student’s Grade According To Their Social Ties

Show me your friends and I’ll tell you your next test score! A recent study conducted by doctoral students at Ben Gurion University shows that there is a strong correlation between a student’s performance in class and the people he or she chooses to study with.

Researchers Michael Fire and Gilad Katz are part of the Data Security on Social Networks research group, which is part of the university’s Information System Engineering unit. The two reconstructed a class’ social network, using information gathered from assignment handed in online (10,759 entries).

Related articles

• Israeli Entrepreneurs Use Facebook To Teach Kids History
• FreezeStick: The Cool Invention By Israeli Kid Scientists

The researchers kept track of students who routinely handed in joint assignments and also monitored which individual assignments were handed in using one computer, indicating that the students worked on them together. Fire tells NoCamels: “We discovered that we can reconstruct the class’ social ties by looking at the students’ hand-in methods and schools website logs.”

Best friends – best indication

When analyzing the data, Fire and Katz discovered that the leading factor for predicting a student’s final test score, was the performance of their “best friend”, meaning the person with whom they’ve had the most interaction on the class’ social network.

In addition, the study has shown a correlation between copying homework and receiving lower test scores.

Fire offers a number of explanations for this outcome: “A student’s friends affect their performance in class – meaning that if you choose your friends wisely, you will get a higher grade.” Alternatively, suggests Fire: “Students’ social networks conform to the assumption that similar people attract each other (Homophily) – meaning better students will study with similar students, and worse students will do the same.”

The research group Katz and Fire belong to have done many other studies in the field of social networks. “We believe that our model can work on different social networks in different scenarios,” Fire tells NoCamels. “In recent years, many studies presented different connections between the individuals in the social networks and their friends.”

The other studies include predicting links between individuals on the same network, identifying fake users on social websites and even social-network-based algorithms for preventing automobile accidents.

The paper, dubbed “Predicting Student Exam’s Scores by Analyzing Social Network Data,” was recently presented at the Active Media Technology conference in Macau, China.

Photo by hackNY