SmartChoice: An Online Recommender System to Support Low‐Income Families in Public School Choice

Abstract

Articles AI MAGAZINE Fami lies whose children attend schools that are not mak ing adequate yearly progress (AYP) under the guidelines of the 2001No Child Left Behind (NCLB) Act are granted the legal right to make an important decision-whether to send their child to a different school. When this occurs, school districts must give students the option of moving to a school that is meeting its AYP goals. However, across the country, fewer than 6 percent of eligible students take advan tage of this provision of the law (GAO 2004, Brown 2004, Howell 2006. Part of the theory of action behind NCLB, and public school choice more generally, is that giving par ents information about school performance and allowing students to leave low-performing schools will improve stu dent outcomes and force poorly performing schools to im prove. However, this provision of NCLB has largely failed to be effective because parents must (1) be able to iden tify which schools will improve their child's performance and (2) actively make choices about their child's education. Research has shown that poor tactical implementations of NCLB policy can effectively strip parents, particularly low-income parents, of their legal options and rights, discourag ing and limiting practical participation and ultimately creat ing a culture of disincentive to exercise choice rights under current policy.Our applied research program examines whether a com prehensive community program, collectively referred to as the SmartChoice Program, centered on a computer-assisted decision support system, SmartChoice, can help to over come this implementation problem in current education pol icy. The decision system provides personalized recommen dations for individual students based on parents' preferences and students' needs,

Incentives in Two-sided Matching Markets with Prediction-enhanced Preference-formation

“…Note that our model is agnostic to where the predictions are formed (e.g., in the school choice setting, it is agnostic to whether a guidance counselor (Nathanson, Corcoran, and Baker-Smith 2013) or a statistical model is making the predictions (Wilson et al 2009)). Previous work shows that not only are algorithms more accurate than humans at making predictions (Dawes, Faust, and Meehl 1989), but also, in some settings, they are also more trusted by participants (Logg, Minson, and Moore 2019).…”

“…However, initially, only a few students took advantage of this opportunity, partly because it was hard for parents to assess which school would improve their child's performance. In 2008, a content-based recommender system called SmartChoice was deployed for focus group participants; its goal was to help parents with the assessment by identifying the best schools based on predictions on the student's development at that school (Wilson et al 2009). Another example is in refugee assignment, where refugees are matched with locations.…”

A MAS Approach to Course Offering Determination

“…Significantly fewer publications are related to the decision-making or administrative tasks in education such as school choice [5], academic advising [6], and course timetable scheduling [7]. The MAS approach has proven an important and effective framework for intelligent educational systems, for example iHelp [8], program planning [9], Time Table Scheduling [10], and personalized study planning [11].…”

“…Individuals have particular preferences, resulting in conflicting goals and therefore leading to conflicts of interest between them. These conflicts should be resolved in a fair cooperative decision making manner; (5). Program administrators consider job markets and the unpredictable nature of preferences students, which generate the need in course scheduling to adapt fast and flexibly to environmental variables and their changes.…”