Virtual advising for high-achieving high school students
GAICIL Insights
The most powerful line in this paper may be a warning.1 Virtual advising for high-achieving high school students, by Oded Gurantz, Matea Pender, Zachary Mabel, Cassandra Larson and Eric Bettinger, tests whether technology can help an overlooked group: high-achieving students from low- and middle-income families who do not apply to the selective colleges where they would likely thrive. Virtual advising, college counseling delivered remotely through technology, was offered to approximately 16,000 such students identified by the College Board, randomly assigned to receive advising from the College Advising Corps.
The results are precise. The offer had no impact on overall college enrollment, yet increased enrollment in high graduation rate colleges by 2.7 percentage points, or 5%, with instrumental variable impacts on treated students of 6.1 percentage points. Demand was strong: 44% of students took up the initiative. Students also followed the specific ideas counselors offered, increasing score sends and enrollment at exactly the colleges identified in the initiative.
Here is the warning. Given how strongly students follow the advice offered, the authors argue, interventions should have very strong priors that the information is actually likely to benefit students. Advice shapes choices, so the quality of advice determines the quality of outcomes. Prior literature supports the view that inducing students to apply to a wider range of more selective colleges is likely to help, which justifies this intervention. Influence and accountability rise together.
Swap the human counselor for an AI advisor and the warning becomes urgent. AI advisors are poised to deliver personalized college and career guidance at a scale human counselors cannot reach. The paper shows that remote, technology-mediated advice can change consequential decisions, and students act on the specific recommendations they receive. An AI advisor would inherit that influence, and the obligation to ensure its recommendations are well-founded. A persuasive machine with mediocre priors is worse than no advisor at all.
The upside is large. Human advising is costly and unevenly distributed, and high-achieving students from lower-income families often lack informed guidance. Virtual delivery already lowers costs. AI could lower them further, offering timely, tailored information to every eligible student. If effects like these could be reproduced at scale, more talented students would enroll in colleges with higher graduation rates, with lasting benefits for them and the economy.
The design rules follow directly. Because students respond to specific suggestions, errors in an AI advisor's recommendations would be acted upon. Designers should ground recommendations in evidence of student benefit, as the authors urge for any intervention. Transparent criteria, rigorous testing and evaluation through randomized trials would help ensure AI advising improves outcomes. An advice engine should be audited like a medical device, since its outputs get swallowed whole.
The null effect on overall enrollment is quietly important. Virtual advising shifted students toward better colleges without changing whether they enrolled. The value of guidance lies in improving the match between students and institutions. AI tools that help students identify colleges where they are likely to succeed could produce similar gains in match quality, a channel that matters for completion and for returns to education. Better matches are the prize.
For policy, the study supports scalable advising, with a premium on evidence. Governments, schools and nonprofits can use technology to extend counseling to students who lack it. As AI tools arrive, these institutions can pair them with human counselors and evaluate effects carefully, ensuring the information serves students' interests.
The thread running through it all: information delivered at the right moment changes life trajectories. Technology makes such information cheap and widely available; AI will amplify that capability. This paper shows the benefits are real and measurable, and that the obligation to offer sound advice grows with the influence of the advisor, human or artificial.
Here is the uncomfortable test for edtech founders: the better an AI advisor is at persuading teenagers, the higher the evidentiary bar it should clear before launch. Compliance is the feature that makes accuracy mandatory.
Introduction
We examine whether virtual advising – college counseling using technology to communicate remotely – increases postsecondary enrollment in selective colleges. We test this approach using a sample of approximately 16,000 high-achieving, low- and middle-income students identified by the College Board and randomly assigned to receive virtual advising from the College Advising Corps. The offer of virtual advising had no impact on overall college enrollment, but increased enrollment in high graduation rate colleges by 2.7 percentage points (5%), with instrumental variable impacts on treated students of 6.1 percentage points.
Takeaway
We find that providing an offer of virtual advising to low- and middle-income, high-achieving students increases enrollment in colleges with higher graduation rates. High-achieving students have strong demand for receiving this additional support, with 44% taking up the advising initiative. Students are also responsive to the specific ideas proffered by counselors, as they increased their score sends and enrollment in exactly those colleges identified in the initiative. Given this observed responsiveness, interventions should have very strong priors that the information being offered is actually likely to benefit students. Overall, inducing students to apply to a wider range of more selective colleges is supported by prior literature as likely to produce positive impacts (Hoekstra, 2009; Zimmerman, 2013).2,31
Future information on degree completion rates for this initiative, as well as other similar projects, will be vital for determining whether encouraging students to attend more selective colleges ultimately improves labor force outcomes commensurate with any changes in student debt. The increases in average selectivity do not appear to coincide with increases in the expected cost of these colleges for low-income students, though these results should be interpreted with caution. Most importantly, we cannot observe what treatment and control group students ultimately pay for colleges, which might be highly variable after taking into account how colleges shape their incoming class and students’ background characteristics.
This may matter as “opt-in” approaches may dramatically lower participation rates (Bergmann, Lasky-Fink, & Rogers, 2019).
Notes
- Oded Gurantz et al., Virtual advising for high-achieving high school students, 75 Econ. Educ. Rev. 101974 (2020), https://doi.org/10.1016/j.econedurev.2020.101974. Takeaway source: 5. Conclusion, reproduced from a free author or repository copy. ↩
- Hoekstra, M. (2009). The Effect of Attending the Flagship State University on Earnings: A DiscontinuityBased Approach. Review of Economics and Statistics, 91(4), 717-724.. ↩
- Zimmerman, J. (2013). School Choice, Opportunity and Access: A Geographic Analysis of Public School Enrollment in New Orleans. Retrieved from 17 Table 1. Descriptive statistics and covariate balance Control group mean Individual characteristics Female 47.4% Parent has bachelor's degree 39.5% White 38.4% Hispanic 17.6% African-American 5.2% Asian 32.7% SAT score 1357 School characteristics School s. ↩
Cite as: Oded Gurantz et al., Virtual advising for high-achieving high school students, 75 Econ. Educ. Rev. 101974 (2020), https://doi.org/10.1016/j.econedurev.2020.101974
Header image: Colouring London at RICS (Lecture Hall).jpg by Paul Downey, CC BY 2.0, via Wikimedia Commons.
.jpg){kind=link}