Post (6):Harnessing Peer Effects with EdTech

Problem Statement –

Most students in developing countries study in schools with limited educational resources and have little learning support outside schools. Inside schools, teachers handling large class sizes (and perhaps many classes) have little time and motivation to offer personalized attention to students. In addition, the majority of students from low socioeconomic families with less educated parents have little educational environment and support at home. Under these circumstances, common sense dictates that we utilize classroom peers, an underutilized resource in the educational ecosystem, to support students.

How can we leverage peers to motivate and help students learn in K-12 schools?

Peer Effects in Education –

Peer effects in education is an extensively researched topic – see the detailed review of these studies in Epple and Romano (2011), Sacerdote (2011), and Bishop (2006). Peer effects are broadly defined as the effect of peer background, behavior, and educational outcomes on students’ educational outcomes (and behavior). There are two ways in which peers affect students’ academic outcomes.

1.???? Peers’ behavior/ability/background can generate positive or negative externality for students. Peers can generate positive externality by directly helping students through collaboration or indirectly helping by asking good questions from teachers, allowing teachers to teach at a brisk pace/higher level, or allowing teachers to give more one-to-one time to students. Peers can also cause negative externality by disrupting the class, forcing teachers to teach at a slower pace/lower level, or wasting teachers’ time maintaining discipline.

2.???? Peers could affect students’ learning efforts and outcomes by influencing social norms. Here also, peers can positively or negatively affect students’ outcomes. For example, students may study when their peers exhibit better work ethics and discipline in class. In contrast, peers could also dissuade students from investing in education by branding hard-working students as nerdy or “uncool.”??

Most studies on primary and secondary schools show that the average quality of peers positively affects students’ achievements - called linear-in-means peer effects (Sacerdote 2011). However, these studies also show substantial nonlinearities in peer effects, i.e., peer effects vary based on student’s and peers’ abilities (Hoxby and Weingarth 2005, Lavy et al. 2009). While many (but not all) studies find that high-ability students benefit from similar-ability peers, some studies have also shown that low-ability students benefit the most from the company of high-ability peers. These findings have led to the educational policy of tracking students in classrooms based on their abilities.

Studies showing the benefit of tracking policy on overall students’ achievement, however, attribute it to the fact that teachers can efficiently teach at the ability levels in classes with students of homogeneous abilities. I agree that teaching efficiency may suffer if students’ abilities in the classroom are highly heterogeneous, e.g., if many students have below-grade-level understanding. However, most classes have significant heterogeneity in students’ abilities even after tracking. Thus, understanding how to harness peer effects across students of heterogeneous abilities is still crucial.

Leveraging Peer Effects in Mixed-Ability Classrooms –

Research shows that students’ future economic and social well-being is determined by their cognitive as well as non-cognitive skills, so we should maximize overall skill formation in classrooms (Heckman 2007 & 2008). Therefore, the right question to answer is: ?

How do we leverage peer effects among students of different abilities to maximize the overall skill formation (cognitive and non-cognitive skills) in classrooms?

Past research doesn’t offer a clear answer to this question. While some studies show that classroom heterogeneity benefits overall student cognitive achievements (Vigdor and Nechyba’s 2007), others have shown the opposite (Duflo et al. 2011).

Besides cognitive achievements, peers’ personality traits (non-cognitive skills) could also affect students’ cognitive achievements. Although cognitive and non-cognitive skills for students may be highly correlated, studies have shown that, after controlling for peers’ cognitive skills, their non-cognitive skills still influence students’ academic achievements. For example, ?Golsteyn et al. (2021) show that students perform better in the presence of persistent peers. Shure (2021) examined the “Big Five” personality traits of peers on students’ cognitive achievements and found a positive effect of peers’ conscientiousness on students’ math and language scores. These findings suggest that mixing high-ability students – presumably with higher conscientiousness and better work ethics – could result in higher average cognitive achievements in classrooms.?

Past research also shows that peers’ personality traits can also affect students’ non-cognitive (character) skills. A large body of research finds strong peer effects in adverse non-cognitive skills, such as drug use, criminal activities, alcohol drinking, smoking, churchgoing, teen pregnancy, and school dropouts (Gaviara and Raphael 2001, Sacerdone 2011 for review of these studies). Xu et al. (2022) show that a higher proportion of repeating students affect nonrepeating students’ cognitive achievements– due to the spillover of poor work ethics from repeating to nonrepeating students. However, little guidance is available on how peers’ desirable non-cognitive skills (such as conscientiousness or extraversion) affect students’ corresponding skills.

To summarize, a significant body of research posits that (1) peers’ cognitive and non-cognitive skills influence students’ cognitive achievement, and (2) their non-cognitive skills affect students’ non-cognitive skills (behavior). This literature provides evidence for both - negative peer externalities from low-ability and disruptive students and positive peer externalities from high-ability and disciplined ones.???

Thus, we require interventions to minimize the negative externality of low-ability (disruptive) peers and maximize the positive externality of high-ability peers in classrooms.

Past research does not answer how to achieve it but provides a direction on how to do it. Researchers neither designed nor observed the peer interactions (or peer groups) inside classrooms in most studies. Thus, their estimated peer effects are based on the organic peer interactions inside classrooms. For example, students may only interact with peers of similar abilities (defeating the very purpose of mixing students of different abilities), high-ability students may look down on and thus demotivate their lower-ability peers, or low-ability students may disrupt class because they lack engagement with studies.??

Instead of relying on organic peer interactions, can we design/create desirable peer interactions among students of different abilities in classrooms? Such a proactive approach can motivate low-ability students to engage with studies (minimize negative peer externalities) and incentivize high-ability students to help their lower-ability peers learn (maximize positive peer externality).

I advocate proactively harnessing peer effects in classrooms with EdTech. Suppose we develop group learning activities in classrooms such that high-ability members are incentivized to share knowledge with their low-ability peers and the low-ability peers to learn from them to achieve the group goals. If students conduct these activities on an EdTech platform, teachers/educators can monitor their interactions and take corrective action in real-time.

Harnessing Peer Effects with EdTech –

I implemented this approach in my peer-driven knowledge diffusion EPInc platform. The platform does the following.

-??????? Identify students’ knowledge gaps – the AI engine identifies students’ knowledge gaps and offers personalized learning content.?

-??????? Peer group manipulation – The platform creates balanced teams by grouping students of different abilities. So, every team has high-ability students to help low-ability peers. Thus, high-ability peers can help students with knowledge gaps and motivate them to study. High-ability peers may also set better group norms for work ethics for their teammates to emulate. ????

-??????? Design, incentivize, and monitor peer-led knowledge diffusion – ?Incentivize high-ability peers to share knowledge with their teammates by designing team-based rewards - winning team-based class tournaments. The social recognition of winning tournaments may motivate students of all abilities to help teammates and contribute to their team performance. Teachers can monitor peer interactions in real-time on the EPInc platform and take corrective actions when needed.??

Besides cognitive skills, while helping teammates and putting in individual efforts to contribute to the team’s success, students may also develop character skills. Specifically, students may build the non-cognitive skills of generosity, trust, reciprocity, and cooperative behavior.?

The details of how the EPInc platform works in classrooms are available in my previous post: https://www.dhirubhai.net/pulse/post-5-epinc-ai-enabled-knowledge-diffusion-platform-anuj-kumar.

Sneak-Peek of the Next Post –

Although the EPInc platform is designed for team members to share knowledge, whether and how they share knowledge is an empirical question to be tested in a field study. In my next post, I will describe how I estimate the effect of peer-driven knowledge diffusion on the EPInc platform on the distribution of cognitive and non-cognitive skills of students in classrooms. Specifically, I will describe the design of a large-scale field experiment on over 2,000 students in 60 3rd to 6th-grade classes in Indian schools.

?

References

Bishop, J. (2006). Drinking from the fountain of knowledge: Student incentive to study and learn–externalities, information problems and peer pressure.?Handbook of the Economics of Education,?2, 909-944.

Duflo, E., Dupas, P., & Kremer, M. (2011). Peer effects, teacher incentives, and the impact of tracking: Evidence from a randomized evaluation in Kenya.?American economic review,?101(5), 1739-1774.

Epple, D., & Romano, R. E. (2011). Peer effects in education: A survey of the theory and evidence. In?Handbook of social economics?(Vol. 1, pp. 1053-1163). North-Holland

Gaviria, A., & Raphael, S. (2001). School-based peer effects and juvenile behavior.?Review of Economics and Statistics,?83(2), 257-268.

Golsteyn, B. H., Non, A., & Z?litz, U. (2021). The impact of peer personality on academic achievement.?Journal of Political Economy,?129(4), 1052-1099.

Heckman, J. J. (2007). The economics, technology, and neuroscience of human capability formation.?Proceedings of the national Academy of Sciences,?104(33), 13250-13255.

Heckman, J. J. (2008). Schools, skills, and synapses.?Economic inquiry,?46(3), 289-324.

Hoxby, C. M., & Weingarth, G. (2005).?Taking race out of the equation: School reassignment and the structure of peer effects?(No. 7867). Working paper.

Lavy, V., Silva, O., & Weinhardt, F. (2009).?The good, the bad and the average: Evidence on the scale and nature of ability peer effects in schools?(No. w15600). National Bureau of Economic Research.

Neidell, M., & Waldfogel, J. (2010). Cognitive and noncognitive peer effects in early education.?The Review of Economics and Statistics,?92(3), 562-576.

Sacerdote, B. (2011). Peer effects in education: How might they work, how big are they and how much do we know thus far? In?Handbook of the Economics of Education?(Vol. 3, pp. 249-277). Elsevier.

Shure, N. (2021). Non-cognitive peer effects in secondary education.?Labour Economics,?73, 102074.

Vigdor, J., & Nechyba, T. (2007). Peer effects in North Carolina public schools.?Schools and the equal opportunity problem, 73-101.

Xu, D., Zhang, Q., & Zhou, X. (2022). The impact of low-ability peers on cognitive and noncognitive outcomes: Random assignment evidence on the effects and operating channels.?Journal of Human Resources,?57(2), 555-596.