Contribution to the body of knowledge

Every doctorate, and every research paper, is judged on what it adds to what was already known. The requirement is stated in regulations but rarely defined in them, and candidates are often left to infer it. The articles below examine the question directly: what counts as a contribution, who decides, and how a candidate demonstrates one.

The first two groups cover what originality means and how examiners judge it, the third treats the question in general terms, and the last is specific to computer science and artificial intelligence. Each entry says briefly why it is worth reading.

• The doctorate as an original contribution to knowledge: considering relationships between originality, creativity, and innovation

  Baptista, A., Frick, L., Holley, K., Remmik, M., Tesch, J., & Åkerlind, G. (2015). Frontline Learning Research, 3(3), 55–67.

  PDF on this site

  Separates three ideas that are often used interchangeably. Originality is required of a doctorate; creativity describes the process that may produce it; innovation concerns its uptake by others. A thesis can be original without being innovative, and the examiner asks only for the first.

• Originality and the PhD: what is it and how can it be demonstrated?

  Gill, P., & Dolan, G. (2015). Nurse Researcher, 22(6), 11–15.

  PDF on this site

  Argues that originality is rarely a wholly new discovery. It is more often a known method applied to a new problem, a new population, new data, or a synthesis that has not been attempted before, and the candidate is expected to state which of these applies.

• What constitutes a “good contribution” to the literature (body of knowledge)?

  Swisher, M. E. University of Florida, Family, Youth and Community Sciences. Teaching handout, not a peer reviewed article.

  PDF on this site

  A practical checklist for judging whether a piece of work adds to the literature. Useful for its directness, though it carries no peer review and should be cited as teaching material rather than as evidence.

• “It’s a PhD, not a Nobel Prize”: how experienced examiners assess research theses

  Mullins, G., & Kiley, M. (2002). Studies in Higher Education, 27(4), 369–386.

  Thirty experienced examiners describe how they judge a thesis. They form an impression early, from the abstract, the structure and the quality of the argument, and they expect a contribution that is sound and defensible rather than spectacular. The source of the page title.

• The concept of ‘originality’ in the Ph.D.: how is it interpreted by examiners?

  Clarke, G., & Lunt, I. (2014). Assessment & Evaluation in Higher Education, 39(7), 803–820.

  Finds that examiners apply the word originality to quite different things, and that the criterion is interpreted more generously in practice than regulations imply. Worth reading before assuming that the examiners share a single definition.

• Doctorateness as a threshold concept

  Trafford, V., & Leshem, S. (2009). Innovations in Education and Teaching International, 46(3), 305–316.

  Treats the ability to argue a contribution as a threshold that a candidate crosses, rather than a section to be written at the end. Explains why a thesis full of competent work can still fail to read as doctoral.

• Threshold concepts in research education and evidence of threshold crossing

  Kiley, M., & Wisker, G. (2009). Higher Education Research & Development, 28(4), 431–441.

  Identifies the conceptual steps candidates must take, among them argument, theorisation and framework, and reports what crossing each one looks like in supervision. A companion to the paper above.

• What constitutes a theoretical contribution?

  Whetten, D. A. (1989). Academy of Management Review, 14(4), 490–495.

  Proposes the what, how, why, and who/where/when test: which factors are included, how they relate, why that relationship holds, and under what conditions. A compact way to check whether a paper advances theory or only applies it.

• Building theory about theory building: what constitutes a theoretical contribution?

  Corley, K. G., & Gioia, D. A. (2011). Academy of Management Review, 36(1), 12–32.

  Revisits the question two decades later and argues a contribution must be original and useful, scientifically or practically. Introduces prescience, the value of anticipating a problem before the field recognises it.

• The changing science of machine learning

  Langley, P. (2011). Machine Learning, 82(3), 275–279.

  A discussion paper on what the field has come to reward. Langley observes that machine learning has moved from the study of learning as a phenomenon towards the incremental improvement of accuracy on shared benchmarks, and that a paper reporting a small gain on a standard dataset is now accepted more readily than one that explains a mechanism or identifies the conditions under which a method works. The distinction between a measurable result and a scientific contribution is set out plainly.

• Machine learning that matters

  Wagstaff, K. L. (2012). Proceedings of the 29th International Conference on Machine Learning (ICML), Edinburgh, Scotland.

  Argues that the contribution claimed by a typical machine learning paper is detached from any consequence outside the benchmark on which it was measured. Three deficiencies are identified: datasets chosen for availability rather than relevance, metrics that are not interpretable in the terms of the originating domain, and results that are never communicated back to that domain. A candidate is asked to state what would change in the world if the result were adopted.

• Research contributions in human–computer interaction

  Wobbrock, J. O., & Kientz, J. A. (2016). Interactions, 23(3), 38–44.

  Names seven kinds of contribution a computing paper can make, including empirical, artifact, methodological, theoretical, dataset, survey and opinion, and states what each must deliver to count. The most useful single page for a candidate deciding which claim their work actually supports.

• HCI research as problem-solving

  Oulasvirta, A., & Hornbæk, K. (2016). Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 4956–4967.

  Recasts the contribution question as problem-solving: a paper is judged by the significance of the problem it takes on, the effectiveness of the solution and the confidence the evidence supports. Argues that empirical, conceptual and constructive work should be held to the same three tests rather than to separate ones.

• On reproducible AI: towards reproducible research, open science, and digital scholarship in AI publications

  Gundersen, O. E., Gil, Y., & Aha, D. W. (2018). AI Magazine, 39(3), 56–68.

  Treats reproducibility as a precondition of a contribution rather than a courtesy to other researchers: a result that cannot be reproduced has not been shown, and therefore adds nothing that can be built upon. The paper specifies what an artificial intelligence publication must document, covering the method, the data, the experimental procedure and the code, and grades degrees of reproducibility accordingly.

• Are we really making much progress? A worrying analysis of recent neural recommendation approaches

  Ferrari Dacrema, M., Cremonesi, P., & Jannach, D. (2019). Proceedings of the 13th ACM Conference on Recommender Systems (RecSys). Best paper award.

  An empirical test of claimed contributions. Of eighteen neural recommendation methods presented at top level conferences, only seven could be reproduced with reasonable effort, and six of those seven were often outperformed by comparatively simple heuristic baselines. The lesson generalises beyond recommendation: a reported improvement is a contribution only if the comparison it rests upon was conducted fairly.

Using these in practice

Two points recur across them. The first is that a contribution must be stated, not left for a reader to infer: examiners and reviewers both report forming a judgement early, from the abstract and the framing, and a contribution they have to reconstruct for themselves is treated as weaker than one that is named. The second is that originality is interpreted by discipline and by individual, so a candidate is better served by saying explicitly which kind of contribution is claimed, whether a new method, a new application, new evidence, a new synthesis or a new theory, than by relying on the word original alone.

The 100 questions for your VIVA page puts the same requirement in the form an examiner uses, and the abstract structure checker reports whether a contribution is signalled at all in an abstract.