Abstract

As artificial intelligence takes on advising functions and automates both the production of student work and employer-side candidate screening, it threatens to hollow out the value of a university degree and widen existing inequalities. This perspective argues that the most valuable asset a university can offer students in a post-AI economy is credible endorsement, the capacity of a trusted faculty member, advisor, or other mentor to vouch with specificity for a student’s character, competence, and potential. Drawing on social capital theory and mentoring research, the essay introduces the concept of a “vouching gap” to describe the growing divide between students who graduate with credible advocates willing to stake their reputations on their behalf and those who do not.

1 Introduction

In “The Second Coming”, W. B. Yeats imagined a world in which “the center cannot hold” and a rough beast slouches toward Bethlehem to be born (Yeats, 1920). Decades later, Joan Didion borrowed that image as the title of her essay collection on 1960s America, a society in which families, norms, and shared narratives were coming apart at once (Didion, 1968). Both writers named a slow unraveling in which familiar institutions lose their grip before anyone can say exactly when the collapse began. Colleges and universities are now facing similar conditions. The normalization of online education, which began as a pandemic-era necessity, exposed the uncomfortable truth that the campus experience can be reduced to a transactional exchange of content for credits. Now, as colleges are increasingly replacing human advisors with chatbots, and students use AI to answer problem sets, draft essays, and write code, the very meaning of a college degree has come into question (Abbas et al., 2024; Freeman, 2025). These problems are compounded on the employer side, where AI scans resumes, filters applicants, and makes hiring recommendations based on keyword matches rather than human judgment.

Left unaddressed, these developments threaten to hollow out the criteria by which institutions decide who to credential and recommend and the criteria by which employers decide who to hire. It is easy to imagine colleges and universities becoming brittle, transactional platforms for content delivery and credentialing, rather than thick communities that can stand behind the people they graduate. This paper reframes the college degree as an act of vouching and examines how institutions can rebuild that commitment through mentoring relationships, transparent assessment, and human-centered hiring partnerships. I extend recent work on human-at-the-helm approaches to student support (Rhodes, in press) by focusing on higher education’s obligations to provide students with credible endorsement and pathways to opportunity in a post-AI labor market.

2 The vouching gap

One of the most valuable currencies of higher education in the post-AI era is the dense web of educated people who can vouch for students. As used here, vouching refers to a specific, reputation-staking act in which a credible person with firsthand knowledge of a student’s work and character publicly endorses that student’s readiness for a particular opportunity, whether a job, a graduate program, or a professional role. Vouching is narrower than mentoring, which encompasses the full range of relational support including emotional care, skill development, and academic guidance. It is also distinct from sponsorship, which typically involves the strategic use of positional power to create opportunity but does not require the deep, sustained observation of character that vouching demands (Hewlett et al., 2010). And it differs from the routine writing of recommendation letters, which often rely on generic language, minimal firsthand knowledge and, increasingly, AI-generated prose. Thus, what makes vouching distinctive is the combination of specificity grounded in direct observation, selectivity that gives the endorsement credibility, and reputational risk on the part of the voucher. In this sense, vouching represents the activation of social capital at the point of institutional transition.

Higher education is uniquely positioned to connect students with credible, high-status professionals who can vouch for their potential. A trusted professor, advisor, graduate student, or other mentor who can send an email or make a phone call, write a specific letter, or provide a personal introduction can be a lifeline. In fact, for many students, these relationships are the primary engine of success. Research consistently shows that such connections predict stronger feelings of belonging, better retention, and improved academic performance (Chetty et al., 2022; Schwartz et al., 2023). Raposa et al. (2021) analyzed data from more than 30,000 college graduates via the Gallup-Purdue Index and found that having someone on campus who cared about students as people and supported them in pursuing their goals was among the strongest predictors of educational success. Similarly, community college students who spoke with faculty about academic matters outside of class showed significant gains in GPA, persistence, degree attainment, and transfer to four-year institutions (Schudde, 2019).

Relationships and connections are also a major factor in job placements. In a landmark five-year experiment involving 20 million LinkedIn users, Rajkumar et al. (2022) found that social network ties, particularly in the form of professional connections, increase job mobility. And, as automation takes over routine tasks, employers are likely to increasingly value interpersonal skills. A trusted mentor’s ability to speak credibly about a student’s character and strengths often becomes the decisive factor in labor market decisions (Deming, 2017). Hagler et al. (2021) found that first-generation students who lacked access to professional networks and trusted advocates faced significant disadvantages in college-to-career transitions because no one with standing in their field was positioned to testify on their behalf. Without intentional mentoring systems, this “vouching gap” will only widen. Economically advantaged students can continue to rely on well-connected networks of friends and family, while their less privileged peers will be increasingly directed to student-facing chatbots that cannot credibly vouch on their behalf (Lai et al., 2025; Zao-Sanders, 2024). This shift deprives those who need it most of the very relationships and personal recommendations that open doors to social mobility.

3 Character, trust, and human presence

But what, exactly, would a mentor vouch for? When credentials can no longer be taken at face value, what remains is character, including honesty, integrity, perseverance, practical wisdom, and concern for the common good, that a credible adult can both model and observe (Immordino-Yang et al., 2019). Recognizing those qualities demands sustained human attention and empathy, the kind that notices when a student is avoiding a difficult challenge or when genuine effort begins to take hold. As Rubin et al. (2024) argued, empathy is conveyed in part by one’s willingness to spend limited cognitive and emotional resources on another person. A chatbot’s attention, by contrast, is essentially cost-free, and it would react with comparable enthusiasm and patience to anyone else. Shared experience also plays a vital role. When a professor reveals their own sophomore-year C in Chemistry, they use this setback to convey genuine understanding and hope. That kind of presence, costly, finite, and grounded in real experience, is what makes mentoring irreplaceable.

The good news is that AI, paradoxically, may create the conditions for the kind of education that makes human connection and vouching more scalable. As AI assumes more of the cognitive load that has traditionally consumed so much of higher education, from memorizing facts to producing first drafts to analyzing routine datasets, it frees up time and institutional energy for apprenticeship-style, experiential learning. When students collaborate with graduate students and faculty members on research projects, work through ethical dilemmas in clinical placements, contribute to community-based projects, or mentor more junior students, they demonstrate, in real time and under observation, the character and qualities that will matter most to employers. This includes qualities that no transcript can certify and that algorithms, even where technically capable of detecting behavioral signals, cannot responsibly assess because such inferences lack the context and personal accountability. It might include intellectual ability and curiosity, honest self-assessment, persistence through difficulty, the capacity to solve problems and collaborate across differences, and the willingness to take responsibility when things go wrong. A mentor’s vouching carries weight because it is selective. The increasingly bland, AI-generated faculty recommendations, often offered indiscriminately, without firsthand knowledge of a student’s work and character, cheapens the currency for everyone. This is why contexts for sustained observation matter so much. Without them, the students who most need credible advocates are the ones least likely to have them. Colleges that take this seriously would move experiential learning from the periphery of the curriculum to its center, not as an elective add-on but as the primary site where students develop and demonstrate the competencies that will define their professional identities.

4 A human-at-the-helm model

The question, then, is how to make this vision work at scale. Here the concept of a human-at-the-helm model can help close the vouching gap. Rather than replacing human support with chatbots, a human-at-the-helm approach positions AI as a cognitive assistant that works behind the scenes to make advising and mentoring more effective and scalable. Before a meeting, AI can prepare on-demand summaries of each student’s goals, challenges, grades, notes, and conversation history, freeing mentors from the cognitive burden of gathering and tracking all this data, and enabling them to be more present and to offer personalized guidance at scale. In doing so, this model allows mentors to reallocate their finite resources away from administrative tracking and ad hoc problem solving toward apprenticeship and empathic care.

The strength of this approach lies in its integration with institutional data and retrieval-augmented generation (RAG) models trained specifically on student success, mental health, workforce development, and more. Unlike generic tools built on static or public information, this system would connect directly to enrollment and degree-audit records while also drawing on specialized knowledge about how students navigate academic and personal challenges. When a student considers dropping a chemistry course, the system immediately surfaces the downstream consequences, whether the withdrawal would jeopardize a pre-med track or lower the student below the credit threshold required for financial aid, and contextualizes these facts using relevant support insights. Advisors carrying large caseloads can arrive at each meeting already equipped with this blend of local institutional data and evidence-based guidance, allowing the meeting to focus on the student’s values and goals rather than on administrative problem-solving. It could also help with some of the drivers of attrition, including student belongingness and mental health. Consider a mentor advising a first-generation student who is doubting whether they belong in a challenging STEM major. Drawing on the student’s local, personalized data and curated evidence-based resources, the system could surface specific strategies from belonging-uncertainty research and “wise” interventions on feedback and belonging (Walton and Cohen, 2011; Yeager et al., 2014). It could also surface single-session interventions (SSIs), structured online activities that take fifteen to thirty minutes and have been shown to improve hope, self-efficacy, depression, and anxiety (Schleider et al., 2021). The mentor stays fully in control, deciding what to say, how to say it, and when to probe further, potentially pushing an ordinary advising session toward a turning point in a student’s development.

5 Discussion

If Yeats (1920) and Didion (1968) give us a language for institutional unmooring, we might counter with the idea of vouching towards Bethlehem, reclaiming the degree not merely as a credential but as a social contract. In an economy where AI can replicate the outputs that once justified credentialing, the signal value of a transcript or diploma is eroding, and what remains is social capital, including the web of relationships, endorsements, and reputational investments that connect graduates to opportunity. The vouching model proposed here is grounded in established theory and growing empirical evidence but realizing it will require both institutional commitment and rigorous empirical testing.

On the institutional side, colleges and universities that orient themselves toward vouching would formalize mentoring as a distinct, evaluable criterion in tenure and promotion decisions, documenting specific metrics such as research opportunities created, graduate school and job placements, personal outreach, and job referrals made. They would also move apprenticeship and experiential learning from the margins to the core of the undergraduate experience, recognizing that these are the settings where character is formed, where vocational skills and identity are developed and demonstrated, and where adults gain the firsthand knowledge of students that makes vouching possible. And they would invest in AI-augmented mentoring systems, built on data and curated, peer-reviewed knowledge bases, that ensure that every student, not just the well-connected, is able to graduate with a range of caring adults who can speak with specificity and confidence about their integrity, their judgment, their persistence, and their readiness to contribute.

Such reforms may face a range of implementation barriers. For example, formalizing mentoring in tenure and promotion decisions would require confronting entrenched incentive structures, where research productivity is often weighted more heavily than teaching and service work in practice, even when written policies suggest otherwise (Dennin et al., 2017). Additionally, the AI-augmented advising infrastructure proposed here will require institutional data integration, governance protocols, and sustained technical support that may be prohibitive at under-resourced colleges and universities. The application and utility of the model may also vary across educational contexts. At community colleges, where heavy advising loads leave little time for the sustained relationships that vouching requires, AI-augmented advising may be the most practical lever for creating that relational space (Hagler et al., 2021; Schudde, 2019). In online programs, where the conditions for sustained observation do not arise naturally, institutions should deliberately build them into the curriculum through research collaborations and capstone projects with named faculty sponsors. In global settings, which are often characterized by high student-to-faculty ratios and exam-based credentialing, AI-augmented advising may similarly help faculty manage larger caseloads while preserving space for the relationships that make endorsement credible (Marginson, 2016).

Ethical questions would also need to be addressed before implementation. Students would need to provide informed consent for the aggregation and use of their academic, behavioral, and personal data, as well as transparency about what information their advisors can access and how AI-generated summaries are produced. Compliance with data privacy regulations, including FERPA in the United States and GDPR for institutions serving students in Europe, would require clear governance protocols regarding data access, retention, and review, and advisors would need to be trained to critically evaluate AI-generated content for potential biases and inaccuracies. Likewise, institutional reforms should not outpace research evidence. Several lines of empirical investigation are needed to determine whether structured vouching actually narrows equity gaps or merely formalizes existing advantages. The most direct test would be a randomized controlled trial in which students are assigned to receive structured mentoring with an explicit vouching component, including portfolio-based endorsement, faculty-initiated introductions to employers, and personalized recommendation protocols, compared with students receiving standard advising. Labor market outcomes including time to first professional employment, starting salary, and alignment with degree field could be assessed post-graduation, stratified by generational status, race and ethnicity, and family income. Likewise, there is a need to compare the human-at-the-helm advising model with both chatbot-only advising and traditional advising without AI support, measuring not only academic outcomes but also advising session quality, student-reported working alliance, and the specificity and behavioral detail contained in subsequent letters of recommendation. Process data from the AI system itself, including which output advisors accept, modify, or reject, could illuminate how human judgment and algorithmic suggestion interact in practice.

Equally important is research on the equity dimensions of vouching itself. Because vouching depends on the formation of close mentoring relationships, it is susceptible to the same homophily biases that shape social networks more broadly (McPherson et al., 2001). In a large-scale audit study, Milkman et al. (2012) found that faculty members were significantly more likely to respond to mentoring requests from white male students than from women or racial minorities, a pattern that was especially pronounced at private institutions and in higher-paying disciplines. If left unstructured, vouching risks replicating these same dynamics, channeling the most credible and consequential endorsements toward students who already hold demographic advantages while leaving others without advocates who carry equivalent standing. Protocols that prompt endorsers to cite specific observed behaviors, projects, and growth over time would make the process less susceptible to the kind of vague, affect-driven assessments that tend to favor students who match a faculty member’s social identity or communication style. Institutions should also track and report vouching outcomes disaggregated by student background, since increased access to support does not automatically translate into equitable outcomes.

Measurement development is needed to advance this line of research. There is currently no validated instrument that captures the construct of vouching as distinct from generic mentoring support or social capital. An instrument that assesses the specificity, credibility, and behavioral grounding of endorsements, as well as the student’s perception that a mentor could and would advocate for them in high-stakes contexts, would allow researchers to test vouching as a mediator between mentoring experiences and labor market outcomes.

Employers who can no longer trust that generic letters of recommendation or that a transcript reflects a student’s capabilities will look instead for someone credible who can say, “I’ve watched this person wrestle with hard problems, work through ethical dilemmas, and show up with honesty and purpose over time, and I’m willing to stake my reputation on their readiness”. That is what vouching means. It is a deeply human act, one rooted in sustained relationships and shared experience. If the center is to hold, higher education must make it the center.

StatementsData availability statement

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

Author contributions

JR: Conceptualization, Writing – original draft, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This work was supported by a grant from Axim Collaborative. The funder had no role in the study design, analysis, decision to publish, or preparation of the manuscript.

Conflict of interest

JR is the co-founder of MentorPRO, an AI-augmented mentoring platform that supports college students and other populations.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

AbbasM.JamF. A.KhanT. I. (2024). Is it harmful or helpful? Examining the causes and consequences of generative AI usage among university students. Int. J. Educ. Technol. High. Educ.21 (10), 10. 10.1186/s41239-024-00444-7

ChettyR.JacksonM. O.KuchlerT.StroebelJ.HendrenN.FlueggeR. B.et al (2022). Social capital I: measurement and associations with economic mobility. Nature608, 108–121. 10.1038/s41586-022-04996-4

DemingD. J. (2017). The growing importance of social skills in the labor market. Q. J. Econ.132 (4), 1593–1640. 10.1093/qje/qjx022

DenninM.SchultzZ. D.FeigA.FinkelsteinN.GreenhootA. F.HildrethM.et al (2017). Aligning practice to policies: changing the culture to recognize and reward teaching at research universities. CBE Life. Sci. Educ.16 (4), es5. 10.1187/cbe.17-02-0032

DidionJ. (1968). Slouching Towards Bethlehem. New York, NY: Farrar, Straus and Giroux.

FreemanJ. (2025). Student Generative AI Survey 2025 (HEPI Policy Note 61). New York, NY: Higher Education Policy Institute. Available online at:https://www.hepi.ac.uk/wp-content/uploads/2025/02/HEPI-Kortext-Student-Generative-AI-Survey-2025.pdf (Accessed May 11, 2026).

HaglerM. A.ChristensenK. M.RhodesJ. E. (2021). A longitudinal investigation of first-generation college students’ mentoring relationships during their transition to higher education. J. Coll. Stud. Retent. Res. Theory Pract.25 (4), 791–819. 10.1177/15210251211022741

HewlettS. A.PerainoK.SherbinL.SumbergK. (2010). The Sponsor Effect: Breaking Through the Last Glass Ceiling. Boston, MA: Harvard Business Review Press.

Immordino-YangM. H.Darling-HammondL.KroneC. R. (2019). Nurturing nature: how brain development is inherently social and emotional, and what this means for education. Educ. Psychol.54 (3), 185–204. 10.1080/00461520.2019.1633924

LaiL.PanY.XuR.JiangY. (2025). Depression and the use of conversational AI for companionship among college students: the mediating role of loneliness and the moderating effects of gender and mind perception. Front. Public Health13, 1580826. 10.3389/fpubh.2025.1580826

MarginsonS. (2016). The worldwide trend to high participation higher education: dynamics of social stratification in inclusive systems. High. Educ.72, 413–434. 10.1007/s10734-016-0016-x

McPhersonM.Smith-LovinL.CookJ. M. (2001). Birds of a feather: homophily in social networks. Annu. Rev. Sociol.27, 415–444. 10.1146/annurev.soc.27.1.415

MilkmanK. L.AkinolaM.ChughD. (2012). Temporal distance and discrimination: an audit correspondence study in academia. Psychol. Sci.23 (7), 710–717. 10.1177/0956797611434539

RajkumarK.Saint-JacquesG.BojinovI.BrynjolfssonE.AralS. (2022). A causal test of the strength of weak ties. Science377 (6612), 1304–1310. 10.1126/science.abl4476

RaposaE. B.HaglerM.LiuD.RhodesJ. (2021). Predictors of close faculty-student relationships and mentorship in higher education: findings from the Gallup-Purdue index. Ann. N. Y. Acad. Sci.1483, 36–49. 10.1111/nyas.14342

RhodesJ. (in press). Human-at-the-Helm: Redefining the role of artificial intelligence in student support. Appl. Dev. Sci.

RubinM.ArnonH.HuppertJ.PerryA. (2024). Considering the role of human empathy in AI-driven therapy. JMIR Ment. Health.11, e56529. 10.2196/56529

SchleiderJ. L.MullarkeyM. C.FoxK. R.DobiasM. L.ShroffA.HartE. A.et al (2021). A randomized trial of online single-session interventions for adolescent depression during COVID-19. Nat. Hum. Behav.6, 258–268. 10.1038/s41562-021-01235-0

SchuddeL. (2019). Short- and long-term impacts of engagement experiences with faculty and peers at community colleges. Rev. High. Educ.42 (2), 385–426. 10.1353/rhe.2019.0001

SchwartzS.ParnesM.BrowneR.AustinL.CarreiroM.RhodesJ.et al (2023). Teaching to fish: impacts of a social capital intervention for college students. Am. Educ. Res. J.60 (5), 986–1022. 10.3102/00028312231181096

WaltonG. M.CohenG. L. (2011). A brief social-belonging intervention improves academic and health outcomes of minority students. Science331 (6023), 1447–1451. 10.1126/science.1198364

YeagerD. S.Purdie-VaughnsV.GarciaJ.ApfelN.BrzustoskiP.MasterA.et al (2014). Breaking the cycle of mistrust: wise interventions to provide critical feedback across the racial divide. J. Exp. Psychol. Gen.143 (2), 804–824. 10.1037/a0033906

YeatsW. B. (1920). “The second coming,” in Michael Robartes and the Dancer, eds. S. Thayer and J.B. Watson (Austin, TX: Cuala Press), p. 466.

Zao-SandersM. (2024). How people are really using GenAI. Harv. Bus. Rev.Available online at:https://hbr.org/2024/03/how-people-are-really-using-genai (Accessed April 27, 2026).

Summary

Keywords

artificial intelligence, equity, higher education, mentoring, social capital

Citation

Rhodes JE (2026) Vouching towards Bethlehem: what colleges and universities owe students in the age of AI. Front. Educ. 11:1855989. doi: 10.3389/feduc.2026.1855989

Updates

Crossmark icon

Check for updates

Copyright

© 2026 Rhodes.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Jean E. Rhodes

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.