Making Assessment Matter
AI as Assessment Partner
Practical Applications and Critical Cautions, Part 1
© 2026 Megan Oakleaf and Becky Croxton
Making Assessment Matter is a five-part C&RL News series focused on maximizing the impact of academic library assessment. The first article introduces strategies for launching assessment projects designed for action and impact. The second explores how centering participants and stakeholders increases the likelihood that assessment results will be used. The third article identifies common pathways for translating assessment results into decisions and action, and the fourth focuses on communicating evidence in ways that resonate with key audiences and support meaningful change. The fifth and final article, published in two parts, examines how artificial intelligence (AI) can thoughtfully support assessment across the entire assessment cycle while keeping expert judgment, ethical practice, and human responsibility at the center. This first part addresses the guardrails necessary for responsible AI use in assessment and the institutional opportunities and risks practitioners should consider. Together, the series equips librarians to use assessment as a driver of learning, decision-making, and continuous improvement.
Introduction
Academic library assessment practitioners have much to consider, starting with the end in mind to ensure assessments lead to action; centering participants to generate meaningful, collaborative evidence; anticipating pathways that connect findings to decisions; and communicating results in ways that reach the right audiences and inspire change. Each of these practices requires knowledge, skill, and professional judgment. Into this established and evolving practice, AI has become impossible to ignore, and library practitioners engaged in assessment work are increasingly confronted with new questions: What guardrails are necessary to ensure that AI supports, rather than undermines, trustworthy, user-centered assessment work? In what ways does it make sense to incorporate AI into library assessment practices across the assessment cycle? And what institutional opportunities and risks should practitioners keep in mind as they incorporate AI into their assessment work? This first part of a two-part article—grounded in the principles, guidance, and competencies outlined in ALA’s draft Guidance on the Use of Artificial Intelligence in Libraries,1 ACRL’s recently approved AI Competencies for Academic Library Workers,2 and ACRL’s Proficiencies for Assessment in Academic Libraries3—addresses the first and third of these questions. The second question, focused on practical AI applications across the assessment cycle, is taken up in part two, appearing in the September 2026 issue of C&RL News.
Essential Guardrails
Current popular and scholarly literature abound with cautions regarding the use of AI and articulation of ethical concerns; four guardrails in particular should be top of mind for practitioners engaged in AI-supported assessment.
- Safeguard assessment data. Among the most obvious guardrails are those associated with data security. To ensure data remain protected and secure, one basic precaution is to limit AI use to only institutionally approved tools.4 Any inclusion of library data in “free” AI options risks sharing protected data outside the institution. Practitioners should also be cautious when considering connecting AI systems to live production databases, lest they introduce data security risks.5 In all cases, practitioners should be aware of institutional regulations and practices governing the use of licensed AI tools and confirm that data-protection safeguards are in place for information exchanged through institutionally approved platforms.6 Of course, practitioners must also carefully follow additional guidelines when conducting any IRB-approved research.
- Maintain critical judgment. Because AI has significant limitations (e.g., it cannot think, feel, or know), a skeptical mindset can serve as a guardrail as well. AI outputs are often sycophantic, misleading, or wrong, and at the same time, they are often grammatically polished, confidently asserted, and superficially authoritative. Practitioners using AI for library assessment work should maintain a critical, evaluative, and discerning stance toward AI outputs rather than accepting them as genuinely authoritative. “[Institutional] systems are built on context, experience, and interpretation;”7 thus assessment practitioners must know their own data intimately in order to accurately evaluate AI results and resist the temptation to turn meaning making over to an AI tool.8 Ultimately, “data, analytics, and AI don’t make decisions, people do.”9 In the case of assessment, practitioners may use AI to more quickly accomplish tasks they already know how to do; if AI introduces misinterpretations, errors, or other concerning content, knowledgeable practitioners will recognize problems and can subsequently choose to course correct or discontinue AI usage for a particular task. Practitioners who experiment with allowing AI tools to assist with tasks they are less familiar with should identify colleagues who can serve as a human interlocutor, recognizing that AI-assisted assessment without human oversight is not responsible practice.
- Honor planning fidelity. When engaging with AI, a practitioner can use assessment planning fidelity as a third important guardrail. AI outputs can beguile even the most sincere assessment practitioner through confident assertions, unsolicited tangents, and plausible-sounding but largely unfounded conclusions while simultaneously tending to agree with users in ways that quietly reinforce confirmation bias, even when prompts explicitly request objectivity. Using an initial research question or user story as a touchstone can help librarians stay focused, resist tangential AI outputs that pull attention away from the original assessment purpose (rabbit holes), and avoid undirected data exploration in search of something interesting to say (data fishing).10
- Document and disclose AI use. Because AI capabilities and usage are advancing quickly, it’s especially necessary to use transparency as a guardrail as well. As standards evolve to expand or limit AI use in assessment, practitioners are obligated to be clear about how their work is augmented by AI usage. Failing to document and disclose use of AI undermines any assessment undertaken. If participants, partners, and target audiences of assessment feel uncertain of whether, how, or to what extent AI was used to support an assessment endeavor, they are likely to (rightfully) question the integrity of the work,11 undermining the trust they have in the process, the results, and the assessment team itself. As Carol Damm and Lance Eaton note, “Documenting without disclosing risks opacity, and disclosing without documenting risks unverifiability.”12 Documentation should include archiving AI interactions, preserving version histories, recording metadata, and ensuring secure storage.13 EDUCAUSE’s Framework for Transparent GenAI Use in Higher Education suggests a four-part framework for disclosing AI use.14 Practices for AI transparency will continue to evolve, but the expectation that AI use in assessment be thoroughly documented and clearly disclosed so results can be questioned, overridden, or reproduced will remain.15
For further exploration of necessary AI-related guardrails, please see ALA’s draft Guidance on the Use of Artificial Intelligence in Academic Libraries.16
These guardrails are essential for responsible, trustworthy AI-supported assessment practice. They shape how assessment work is perceived, valued, and used, both within the library and across the institution. Library assessment practitioners must also consider how their AI use positions the library within broader institutional efforts that are already being augmented by AI-informed decision-making. The implications extend beyond individual projects to questions of institutional relevance and participation, which are explored in the following section.
|
What might this look like in practice? A librarian seeking to process focus group transcripts to identify trends in discussion should not rely on AI for thematic analysis without first working through their own interpretative framework. Although the AI should be able to identify patterns in the transcripts, it should not be trusted to unilaterally ascribe meaning or identify which anomalies or outliers are significant. AI tools cannot know everything a librarian may know about local context, stakeholder groups, individual perspectives, and voices that are often marginalized or omitted. Even with skilled prompting, AI may miss moments of insight that a human instinctively recognizes as meaningful. And if AI were to someday achieve this level of expertise, librarians should still not rely on AI agents to work independently, if for no other reason than doing so would likely limit the depth of understanding needed to clearly articulate the work and its implications to colleagues, stakeholders, and impacted groups and individuals. |
Institutional Opportunities and Risks
Academic libraries are inextricably linked to their overarching institutions, so any consideration of AI-supported library assessment must be considered within both institutional and library contexts. Certainly, academic library assessment centers many of the same foci and values that higher education assessment prioritizes, including student learning, success, and experience; teaching excellence encompassing curricula, pedagogy, and learning assessment; and research productivity and funding. This emphasis is reinforced in the recently revised and approved (April 2026) ACRL Standards for Libraries in Higher Education, which highlight the importance of aligning outcomes, building evidence of impact, and communicating the library’s role in student learning and institutional success.17 Library assessment intersects other areas of institutional interest as well, including admissions and affordability, rankings and prestige, post-graduation outcomes, global and community impacts, and so on. Increasingly, higher education institutions are leveraging AI to analyze and assess qualitative and quantitative data in all these areas. The growing use of AI at the campus level enables institutional leaders to track fast-paced changes at their institutions specifically and in higher education more generally; ask and answer more questions; identify areas of excellence or vulnerability; and make decisions and take action more quickly. The scale and pace of AI use for these purposes at an institutional level is rapidly transforming how thinking and decision-making occur in the academy.
In the context of higher education AI-supported assessment, librarians, assessment practitioners, and library leaders have new opportunities to engage in institutional conversations, though they risk being left out if they choose not to participate. Assessment practitioners who are attuned to AI-informed campus conversations and initiatives are better situated to contribute to institutionally important discussions and problem-solving efforts. Those who go a step further by engaging in these conversations—either by contributing library data to institutional analyses to create more complete pictures of student and faculty experiences or by conducting AI-supported library assessments that complement institution-level analyses—can position the library as essential to institutional success.
Beyond institutional positioning, those who engage with AI-informed approaches have an opportunity to shape what responsible, equity-conscious assessment looks like in higher education. Assessment practitioners have a professional commitment to user-centered inquiry, a skepticism toward data that privileges “average” and flattens complexity, and a habit of asking whose perspectives are represented and whose are missing. Indeed, library assessment practitioners are among the best-prepared professionals to critically evaluate AI-informed assessments, advocate for their thoughtful use, and model expert AI literacy skills for their institutional colleagues. At this moment, a window of opportunity exists for libraries to shape campus use of AI-informed analyses. In the years to come, librarians will be needed to continue this work as technologies proliferate. Librarians who seek to engage with their institutions might begin by identifying the questions their institutional research office, provost’s office, or similar units are using AI-supported analysis to investigate and determine whether and/or how library data might augment the work, contribute to more complete findings, and lead to enhanced library-campus partnerships. At an institutional level, libraries may also share their data and assessment expertise as campus AI literacy, governance, and compliance evolve.
Conversely, libraries that opt not to contribute library data to institutional AI analyses or use AI in their own library-controlled assessment endeavors risk being omitted from institutional assessment, decision-making, and action-taking in areas central to library missions and institutional success. Over time, this exclusion may marginalize the library and perceptions of its relevance in areas that have long been core library contributions. Institutions that do not include library data in their campuswide analyses risk having a skewed or incomplete picture of areas of institutional importance.
Libraries that choose to contribute data to institutional AI-supported analyses can benefit from productive partnerships with library vendors. Vendors often possess substantial data about engagement with resources and services, which can be significant for building a complete picture of student and faculty experiences. Libraries are best positioned to contribute meaningfully to institutional analyses when vendors share relevant data in accessible formats and support integration with institutional systems using open interoperability standards. In turn, libraries can advise their vendor partners on deidentification practices that align with library and institutional privacy policies and inform other AI-related decisions affecting vendor-supplied library resources and services, including tailoring vendor contributions to reflect institution-specific contexts and governance expectations.18 Vendors who collaborate with the library in these endeavors can be genuine partners in better understanding and supporting their libraries and the institutions of which they are a part.
Conclusion
Together, the guardrails and institutional considerations outlined here provide a foundation for AI-supported assessment that is both responsible and strategically positioned. Practitioners who approach AI with data security, critical judgment, planning fidelity, and transparency in mind are better prepared to engage with it as thinking partners that augment rather than displace the professional judgment, contextual knowledge, and ethical grounding that assessment requires. Libraries that engage with these institutional opportunities from a foundation of responsible practice are better positioned to contribute to campus analyses, shape emerging AI governance conversations, and resist the risk of being sidelined from decisions central to library missions. Part two of this article, appearing in the September 2026 issue of C&RL News, builds on this foundation by walking through practical AI applications at each phase of the assessment cycle. 
Notes
1. “ALA Draft Guidance on the Use of Artificial Intelligence in Libraries Comments Due May 8,” ACRL Insider, April 26, 2026, https://acrl.ala.org/acrlinsider/ala-draft-guidance-of-the-use-of-artificial-intelligence-in-libraries-comments-due-may-8/.
2. Association of College & Research Libraries (ACRL), “AI Competencies for Academic Library Workers,” approved October 2025, https://www.ala.org/acrl/standards/ai.
3. Association of College & Research Libraries (ACRL), “Proficiencies for Assessment in Academic Libraries,” revised and approved June 2023, https://www.ala.org/acrl/standards/assessment_proficiencies.
4. “ALA Draft Guidance - Section 2.3 - Privacy (Principle 1 In Practice).”
5. Bharathwaj Vijayakumar and Samyukta Alapati, “When AI Meets Data: The Promise and the Pressure of Bringing AI into Higher Education Systems,” EDUCAUSE Review, January 29, 2026, https://er.educause.edu/articles/2026/1/when-ai-meets-data-the-promise-and-the-pressure-of-bringing-ai-into-higher-education-systems.
6. ACRL, “AI Competencies - 2.5. Understand AI-related policies and regulations relevant to your context.”
7. Vijayakumar and Alapati, “When AI Meets Data.”
8. “ALA Draft Guidance - Section 2.1 Public Good (Principle 1 In Practice); Preserving Human Decision-Making, Self-Determination, and Labor.”
9. Deborah Dent, James Frazee, Carrie Shumaker and Tim Wrye, “2025 EDUCAUSE Top 10 #1: The Data-Empowered Institution,” EDUCAUSE Review, October 23, 2024, https://er.educause.edu/articles/2024/10/2025-educause-top-10-1-the—data-empowered—institution.
10. ACRL, “Proficiencies for Assessment - 3. Designing, Collecting, and Analyzing, Item 3.3.”
11. Carol Damm and Lance Eaton, “From Prompt to Practice: A Framework for Transparent Gen AI Use in Higher Education,” EDUCAUSE Review, March 18, 2026, https://er.educause.edu/articles/2026/3/from-prompt-to-practice-a-framework-for-transparent-genai-use-in-higher-education.
12. Damm and Eaton, “From Prompt to Practice.”
13. Damm and Eaton, “From Prompt to Practice.”
14. Damm and Eaton, “From Prompt to Practice.”
15. “ALA Draft Guidance - Section 4.4 Responsibility & Explainability.”
16. “ALA Draft Guidance.”
17. Association of College & Research Libraries (ACRL), “Standards for Libraries in Higher Education,” revised and approved April 2026, https://www.ala.org/acrl/standards/standardslibraries.
18. Vijayakumar and Alapati, “When AI Meets Data.”
Article Views (By Year/Month)
| 2026 |
| January: 0 |
| February: 0 |
| March: 0 |
| April: 0 |
| May: 0 |
| June: 9 |
| July: 1087 |