If you’ve followed the AI-in-education conversation at all, you’ve heard the word “originality” thrown around constantly. Universities build entire policies around it. Plagiarism detection tools claim to measure it. Students are told their work must be “original” or face consequences. But what does originality actually mean in academic research? The answer, it turns out, is far messier than most policies assume.
Guetzkow, Lamont, and Mallard (2004) interviewed 49 peer-review panelists across five prestigious multidisciplinary fellowship competitions in the humanities and social sciences. These weren’t casual conversations. The panelists were experienced scholars tasked with evaluating the best proposals in their fields, and the researchers wanted to know how they defined originality when real decisions were on the table.
The traditional view, rooted in Merton and Kuhn’s sociology of the natural sciences, reduces originality to two things: new discoveries and new theories. Guetzkow et al. (2004) found that this framework collapses when applied outside the natural sciences. Their panelists described seven broad categories of originality: original approach, original topic, original theory, original method, original data, original results, and research in an understudied area. The range alone tells you something. Originality isn’t one thing. It never was.
Original Approach Tops the List
The single most frequently mentioned form of originality was “original approach,” accounting for nearly a third of all references. Panelists described it using words like “fresh perspective,” “new angle,” “new questions,” and “new framing.” Guetzkow et al. (2004) note that this category doesn’t exist in the canonical sociology of science literature, and it operates at a higher level of generality than specific theories or methods. It’s about how you construct a problem, not the tools you use to study it.
And here’s what makes this so relevant to the current AI debate. “Original results,” the form of originality that dominates natural science discussions, was the least mentioned category among these panelists. Only 4% of originality references pointed to results. Even when results came up, panelists talked about “new interpretations” more often than new discoveries.
This should give pause to anyone writing AI policies that define originality in terms of output. I covered Luo’s (2024) review of how universities tried to build originality policies around generative AI, and found them inconsistent, vague, and often contradictory. Part of the problem is clear now: the policies are working with a narrow definition of originality that doesn’t match how scholars actually evaluate academic work.
Disciplinary Patterns
Guetzkow et al. (2004) found clear disciplinary differences in how panelists valued different types of originality. Humanists and historians prized original approach above all else, with 33% and 43% of their mentions respectively. Humanists also placed high value on original data, meaning new texts, archives, photographs, musical scores, and similar materials.
Social scientists showed a more diverse spread. They most often valued original methods (27% of mentions) but also appreciated original approach, theory, and topic in roughly equal measure. The picture that emerges is one of intellectual diversity: different disciplines don’t just study different things, they have fundamentally different ideas about what counts as new.
This has real implications for how we talk about AI and student work. When a university policy says students must submit “original” work and not rely on AI-generated content, which definition of originality is it using? If originality means a fresh approach to constructing a problem, then a student who uses AI to handle routine writing but brings a genuinely new angle to their analysis might be producing more original work than someone who writes every word themselves but rehashes conventional arguments.
Chan (2023) explored this tension in his study of student perceptions of AI misconduct and AI-giarism, and found that students couldn’t agree on where the line was. Guetzkow et al.’s data, collected twenty years earlier, helps explain why: the concept itself has always been contested.
The Moral Dimension of Originality
This is where the paper gets really interesting. Guetzkow et al. (2004) found that panelists associated originality with the moral character of the researcher in about 40% of their discussions. Original scholars were described as:
Applicants’ whose proposals were deemed original were often described with such adjectives as adventurous, ambitious, bold, courageous, curious, independent, intellectually honest, and risk-taking. They were also viewed as “going out of their way,” “challenging the status quo,” “thinking for themselves” and “having a passion for ideas.” (p. 203)
Unoriginal scholars got labeled:
Unmotivated or incapable of independent thought and were described with terms that include: conformist, complacent, derivative, facile, gap-filling, hackneyed, lazy, parochial, pedestrian, rehashing, tired, traditional, uncritical, “spinning their wheels;” or alternatively, fashionable, trendy, “shambolic,” slavish, “riding on the band wagon” or “throwing around buzz words.(p. 203)
A full 81% of panelists made at least one connection between originality and moral qualities. The core virtue being assessed was intellectual authenticity. Panelists rewarded scholars who seemed to follow genuine intellectual interests regardless of professional risk. They penalized those who appeared to reproduce their advisor’s work, chase trends, or, as one panelist put it, “throw around buzz words.”
Guetzkow et al. (2004) explain the logic:
The problem with trendiness, as with reproducing the status quo, is not so much that it reflects conformity, laziness, dishonesty, faddishness or disingenuousness. It is that these qualities characterize people who are inauthentic, or scholars who lack “genuine” intellectual interests and passion. Otherwise, they would break free of the inertial forces of academia to pursue their ideas at all costs, instead of just going with the flow (p. 205).
I find this finding both illuminating and troubling. It’s illuminating because it names what a lot of faculty are really worried about with AI: not that students are getting answers wrong, but that they’re not thinking for themselves. The fear isn’t plagiarism. It’s intellectual outsourcing. Dawson, Bearman, Dollinger, and Boud (2024) made a similar argument when they said our obsession with catching cheating has overtaken our concern for whether assessments are actually valid.
But it’s also troubling because moral judgments about a researcher’s character introduce enormous subjectivity into peer review. The line between “courageous” and “reckless,” between “bold” and “unfocused,” is in the eye of the reviewer. Guetzkow et al. (2004) acknowledge this: the peer review literature tends to draw a sharp line between “legitimate” substantive evaluations and “illegitimate” personal considerations, but their data shows these categories are tangled together in real evaluations.
A 2004 Paper That Speaks to 2026
This paper is over two decades old, and it doesn’t mention AI at all. But the questions it raises about originality are exactly the ones we’re struggling with now. If originality means a fresh approach, intellectual authenticity, and the courage to ask new questions, then the conversation about AI and academic work needs to move beyond “did the student write this themselves?” That question, by Guetzkow et al.’s own framework, targets the least important form of originality.
The authors themselves called for a broader view: “A widespread practice of taking the natural sciences as a normative model has generated problems in the theoretical cultures of the social sciences in particular” (p. 206). Twenty-two years later, the same critique applies to AI policy. We’re using a narrow, output-focused definition of originality to govern a complex, multi-dimensional concept. And the mismatch is creating policies that miss the point entirely.
References
- Chan, C. K. Y. (2023). Is AI changing the rules of academic misconduct? An in-depth look at students’ perceptions of ‘AI-giarism.’ International Journal of Educational Technology in Higher Education, 20(1), 60. https://doi.org/10.1186/s41239-023-00408-z
- Dawson, P., Bearman, M., Dollinger, M., & Boud, D. (2024). Validity matters more than cheating. Assessment & Evaluation in Higher Education, 49(7), 1005–1016. https://doi.org/10.1080/02602938.2024.2386662
- Guetzkow, J., Lamont, M., & Mallard, G. (2004). What is originality in the humanities and the social sciences? American Sociological Review, 69(2), 190–212. http://www.jstor.org/stable/3593084
- Luo, J. (2024). A critical review of GenAI policies in higher education assessment: A call to reconsider the “originality” of students’ work. Assessment & Evaluation in Higher Education, 49(5), 651-664. https://doi.org/10.1080/02602938.2024.2309963