How do youths learn altruism and equity? Estimating the impact of embedded ethics in education on social preferences

The pre-coded lessons follow a culturally adapted narrative in which the agents play interaction games in which one agent makes a proposal to allocate limited resources for two competing objectives. Throughout the intervention, across 4 to 5 classes, students are assigned to one of the two versions of the pre-coded lesson plan.

2022-10-05

2023-02-28

Our main families of outcomes are (1) responses on the three social interactions; (2) responses on the dictator game and equity game; (3) responses on interaction-related psychosocial wellbeing (midline/endline only).

(1) Three social interactions are ultimatum game and variations of the ultimatum game; (2) we follow the classic dictator game and dictator-equity game used by Rao (2019); (3) for each social-interaction outcome, we ask one follow-up question adapted from the Students’ stress rating scale used in Balamurugan & Kumaran (2008), asking students’ feelings about the outcomes of each interaction on the previous survey.

Our auxiliary families of outcomes include self-reported measures of (4) social preferences in respect to altruism/equity/disability inclusion; (5) general psychosocial wellbeing; (6) STEM attitudes; (7) gender attitudes; (8) frequency of blood donation; (9) social attitude; (10) the level of digital exposure.

(4) 15 Questions are adopted from the World Value Survey (2022) and Afrobarometer (2021), and Olaleye et al. (2012) in order to measure social preferences in respect to altruism/equity/disability inclusion; (5) We construct a psychosocial well-being index by adopting questions from Cohen's Perceived Stress Scale (1983) and Diener's Satisfaction With Life Scale (1985); (6) We construct a STEM interests index through questions adapted from the Career Aspiration Scale used in Gregor and O'Brien (2015); (7) The gender attitudes index is adapted from the similar index used in Ahn et al. (2022); (10) We measure the level of digital exposure through the framework of DQ (Digital Intelligence) used to measure literacy and skills.

This current study uses a randomized trial to study the impacts of the embedded-ethics software coding education program. Throughout the intervention, across 4 to 5 an-hour-long extracurricular classes, students are assigned to two treatment groups and a control group.

For one treatment group of students, they are assigned to a version of the pre-coded lesson plan through which they learn how to code decision-making processes of SN (Standard Nash) agents that are interested only in their own payoffs. For another treatment group of students, they are assigned to a different version whereby they learn to code decision-making processes of AE (Altruistic-Equitable) agents, whose interests include payoffs to others. The two versions of the lesson plan share the common structure of containing three sets of game: i) two-player proposal game; ii) three-player proposal game that involves a bystander who receives the benefit equivalent to that of the responder (representing the concept of negative externality); and iii) three-player proposal game that involves a bystander who receives the benefit equivalent to that of the proposer (representing the concept of positive externality). Students learn how to simulate the decision-making logic in the assigned lesson plan by answering questions that review the algebra behind the decision-making processes of both proposer and responder, and by using block-based coding on the Scratch platform to code and simulate such interactions.
As for the data collection instruments, trained enumerators conduct surveys to collect participants’ demographic information, their responses to the pre-set ultimatum game exercises, and their answers to a set of questionnaires examining their interest in STEM, gender attitude, social preferences with respect to altruism, equity, and inclusion, and their psychosocial well-being. Along with the survey, students are also asked questions on activities with real consequences, such as making allocation decisions out of survey-provided funds (specifically designated for the purpose of the exercise) to financially support households with under-resourced children or children with disability. The games and the surveys are conducted three times, once at the beginning of the intervention, once immediately after, and one a few months after the conclusion of the intervention.

Randomization done in office by a computer

Individual

No

Individual randomization

211 students

85 control, 62 Standard Nash, 64 Altruistic-Equitable

The MDE for our main outcome is expected to be between 0.464 standard-deviation units (without controls) and 0.325 standard-deviation unites (assuming controls absorb 30% of residual variation).