AI & Qualitative Research Blog

Cornell University · Malcolm Knowles Award 2023

Annalisa Raymer

Senior Lecturer & Director of CLASP
Cornell University

Dr. Raymer uploaded the transcript of an in-house CLASP film into Evidano to summarize four stakeholder groups’ perspectives and run a thematic analysis, then compared the AI’s reading with feedback from adult-education colleagues at the AAACE conference.

“With Evidano, I feel that I have a straightforward thinking partner, rather than an internet pet eager to please.”

Published in the Journal of Experimental Criminology

Gordon Abner

Associate Professor, LBJ School of Public Affairs
University of Texas at Austin

After the first author completed an inductive thematic analysis of 224 open-ended survey responses on police chief credentials, Professor Abner’s team ran an independent Evidano analysis to check the human coding for oversights, alternative frames, and researcher bias — and weighed, rather than adopted, the extra themes it proposed.

“We found that AI can strengthen qualitative research by providing an independent check on human coding, but final decisions about theme validity should remain with researchers, who must assess whether a theme is sufficiently coherent, frequent, and supported by the data.”

Doctoral dissertation published by UCLA

Johnnie Yang

Postdoctoral Fellow, Semel Healthy Campus Initiative Center
University of California, Los Angeles

For a convergent mixed-methods dissertation on the student wellness crisis, Johnnie Yang used Evidano to run three analyses of student journals and interview transcripts — journals alone, interviews alone, and a combined dataset — then re-read every transcript and made all final interpretations through reflexive thematic analysis.

“Evidano came in on the qualitative side, helping me organize and get an initial read on a huge amount of journal and interview data. I was also the instructor for these students, so I already had relationships and instincts going in, which meant I couldn’t just take the tool’s output at face value. I went back to every transcript myself, checked it against what I actually knew about these students, looked for the stuff that didn’t fit the pattern.”

Published in PLOS Neglected Tropical Diseases

Chisoni Mumba

Associate Professor of Animal Health Economics
University of Zambia

In a hybrid qualitative study of six focus groups near Zambia’s game management areas, Professor Mumba’s team used Evidano to run sentiment analysis on the transcripts, then checked every theme it surfaced against the original transcripts and verbatim quotes.

“Evidano gave us a layer of insight we would have struggled to capture through manual coding alone. It picked up the emotional tone running through the transcripts, the anxiety around eating meat that might be infected, the fear tied to livestock losses, and the frustration participants expressed toward formal health campaigns.”

Published in Studies in Educational Evaluation

Lynette Pretorius

Senior Lecturer, School of Learning and Teaching
Monash University, Australia

After completing reflexive thematic analysis of open-ended survey responses from 29 PhD students and their own editor reflections, Dr. Pretorius and Dr. Cahusac de Caux trialled Evidano for a different perspective on the key ideas in those reflections — and reported the overlap with their constructed themes as evidence for the trustworthiness of their coding strategy.

“AI tools can support your research, but they are not there to do the research for you. In this study, we tested Evidano to illuminate any unconscious biases we may have had about our data. As qualitative researchers, we must stay deeply engaged with our data, and we are responsible for constructing the themes that are included in the final paper.”

Published by Springer at XP 2026

Ehikioya Obode

Researcher, College of Science & Engineering
University of Glasgow, United Kingdom

After completing the thematic analysis of 16 agile-coach interviews manually, Ehikioya Obode’s team uploaded the anonymized transcripts to Evidano for an independent thematic overview, then compared its groupings with their own and returned to the transcripts wherever the two differed.

“Evidano gave us an independent thematic overview after the manual analysis was complete. We did not use it to determine the final themes; instead, we compared its groupings with our own and returned to the transcripts where there were differences.”

Editor, Gifted Child Quarterly

Jennifer Riedl Cross

Editor of Gifted Child Quarterly & Research Professor
University of Louisiana at Lafayette

Facing more than 19,000 open-ended comments from a Mensa Foundation study of 3,443 highly intelligent adults, her team ran the responses through Evidano in batches, compared repeated analyses, and manually verified every theme and supporting quote before finalizing the findings.

“Researchers can work together with Evidano, but they have a responsibility to stay in touch with the data and take steps to confirm the output is accurate and unbiased.”

Published in the Journal of Education Research

Cheng-Ta Wu

Distinguished Professor, College of Education
National Chengchi University, Taiwan

His team’s published article is itself a case study of Evidano: the authors demonstrate its full qualitative workflow — summaries, themes, codebook, frequency analysis, and key quotes — on elementary-school leadership interview data. They propose a human–AI collaborative model for qualitative research.

“Evidano provided transparent traceability between extracted themes, viewpoint-driven queries, and source transcripts, which greatly streamlined the early, laborious stages of coding and quote mapping.”

Published in Educational Psychology

Siv M. Gamlem

Professor of Education
Volda University College, Norway

After two researchers completed the initial inductive thematic analysis of 11 interviews, Professor Gamlem’s team ran the anonymised transcripts through Evidano and compared the results with their manual analysis for triangulation and validation.

“Evidano supported our research process by offering an additional lens on the interview data, helping us identify patterns that could be compared with our initial manual analysis. Its value was not in replacing interpretation, but in strengthening the transparency and robustness of our analytic decisions.”

Published in Cambridge Prisms: Plastics

Adriana Cunha Neves

Researcher, Department of Applied Science
South East Technological University, Ireland

Her team used Evidano for AI-assisted thematic analysis of focus group discussions about starch-protein bioplastics, helping mitigate unintentional researcher bias during interpretation.

“Evidano is particularly valuable for researchers who lack extensive experience in qualitative data analysis, offering an efficient way to organize large datasets while minimizing the influence of pre-existing knowledge and researcher bias.”

Published at IEEE ACDSA 2026

Munn Hong Lam

Researcher, Faculty of Accountancy and Management
Universiti Tunku Abdul Rahman, Malaysia

In a sequential mixed-methods pilot on AI literacy and employment prospects, his team entered the qualitative interview data into Evidano and interpreted it thematically to add depth to their survey results.

“The qualitative element of our research was supported by Evidano through assistance with thematic organisation and interpretation of interview data. This was helpful in our research workflow, particularly in identifying patterns within the responses received during the interviews.”

Published in Annals of Medicine and Medical Sciences

Roland Joseph D. Tan

Ophthalmologist & Health Professions Educator
University of the Philippines Manila

After coding semi-structured interviews and document reviews manually through reflexive thematic analysis, Dr. Tan ran the transcripts through Evidano for a second pass, which produced a complementary codebook and 11 granular themes that he compared with his own analysis before integrating both into the final thematic structure.

“Researchers need to verify AI-generated interpretations against the original data, remain alert to bias and loss of contextual nuance, protect participant confidentiality, and transparently disclose how AI was used. AI should support the researcher’s analytical process, not become the researcher.”

Published in Education Sciences

Shawn S. Savage

Assistant Professor
University of North Carolina Wilmington

After multiple rounds of human coding of narrative student evaluations, Dr. Savage and Dr. Parker used Evidano to interrogate the data from a different analytic vantage point, surfacing gendered comparative distinctions they then verified through collaborative analysis.

“Evidano was most helpful because it encouraged Dr. Parker and me to revisit the data from a different analytic vantage point. It didn’t replace our interpretation; it expanded the kinds of questions we asked.”

Published in Frontiers in Education

Lawrence Meda

Director of Research & Associate Professor
Sharjah Education Academy, UAE

After translating, transcribing, and manually coding nine teacher interviews, Professor Meda’s team used Evidano to categorize participant quotes into researcher-defined themes — with every allocation manually verified by the research team.

“AI-supported tools like Evidano can enhance qualitative research when used carefully and transparently.”

Published in Public Personnel Management

Geneviève Morin

Postdoctoral Researcher
École nationale d’administration publique

To study how U.S. federal employees reacted to the January 2025 return-to-office mandate, Dr. Morin’s team imported 39,788 open-ended comments from three national survey waves into Evidano, coded them against a predefined thematic grid, and manually verified the output before drawing any conclusions.

“AI-assisted coding is a starting point, not a substitute for judgment. We validated every theme against the original comments ourselves before drawing any conclusions.”

Published in Policing: An International Journal

Vincent Mousseau

Professor of Forensic Science
Université du Québec à Trois-Rivières, Canada

For a scoping review covering 25 years of research on civilian police staff, Professor Mousseau’s team used Evidano to structure data extraction from 60 studies around a predefined coding grid, with two coders independently checking flagged studies against the original material.

“I believe researchers using AI-assisted analysis should first remain focused on their research goals: is AI actually helping to achieve those goals, or is it simply being used because it is a new and attractive technology? AI outputs should also always be checked against the original data.”

Published by CICAP, University of Costa Rica

Esteban O. Mora-Martínez

Lecturer & Head of the Evaluation Section
University of Costa Rica

For a book chapter meta-evaluating Costa Rica’s National Evaluation Agenda, Dr. Mora-Martínez used Evidano to apply a single evaluation framework across ten reports spanning employment, education, health, culture, and the environment, acting as evaluator while Evidano analyzed the documents.

“Evidano helped apply the same analytical framework consistently across a diverse set of evaluation reports, making it easier to organize evidence, identify recurring patterns, and compare findings across sectors. At the same time, evaluators should remember that AI does not replace professional judgment.”

Published in Three TESOL Treats (TESOL-Hungary)

Zsuzsanna Soproni

College Professor of English and Research Methods
International Business School, Budapest

After coding 229 Hungarian-language comments from 31 adult sign language learners with the constant comparative method and an independent co-coder, Professor Soproni ran one set of open-ended responses through Evidano as an exploratory check — and found the five themes it proposed pointed to the same priority she had identified by hand.

“It was the first time I’d used an AI tool for thematic analysis, and I was genuinely surprised by how well it handled Hungarian. At the same time, researchers need to remember that while AI can speed up and support the analysis process, it still needs careful human oversight.”

Published in Revista de Cercetare si Interventie Sociala

Yuliia Maslova

Professor & Head of Journalism and PR Management
National University of Ostroh Academy, Ukraine

Building a conceptual model that links fragmented digital identity with soft power, Professor Maslova’s team processed a literature sample of 51 sources with Evidano to surface recurring themes and candidate categories, then verified and interpreted every category manually during theory construction.

“Evidano helped us identify recurring themes and formulate candidate categories across a corpus of 51 literature sources. Its function was supportive rather than substitutive. The authors verified the outputs, interpreted the categories within the theoretical framework, and retained responsibility for every conceptual relation and conclusion.”

Published in Social Sciences & Humanities Open

Enala Sharon Lufungulo

Lecturer & Researcher, School of Education
University of Zambia

In a multi-method study of OER adoption across Lusaka and Chipata, Dr. Lufungulo used Evidano for document analysis, summarisation, and the initial coding of interview transcripts, then cross-checked every output against the original data, her manual coding in Microsoft Word and MAXQDA, and the study’s theoretical framework.

“I also used Evidano as a critical sounding board during the meaning-making process. I compared its proposed interpretations with my own reflections, assumptions and deductions, which helped me consider alternative meanings and critically examine my emerging interpretations.”

Doctoral dissertation published by the University of La Verne

Allison Leanos

Adjunct Professor, LaFetra College of Education
University of La Verne

For a descriptive phenomenological dissertation on how neurodivergent adults experience workplace inclusion and belonging initiatives, Dr. Leanos used Evidano to apply the Moustakas method to ten interview transcripts and generate initial codes and theme concepts, then cross-referenced the codebook against the transcripts and revised the meaning units by hand.

“Nowadays, it’s difficult to avoid AI — it’s everywhere. However, AI is not a properly trained researcher, so it’s important to use it responsibly. Using Evidano helped generate initial codes and theme concepts based on the Moustakas method, which I found helpful and time-effective. I then dissected and interpreted the data manually, which enhances the validity of the results.”

Published in Advance Social Science Archive Journal

Mehak Afzal

MPhil Applied Linguistics Scholar
GIFT University, Pakistan

For a qualitative study of parents of autistic children in Gujranwala, Pakistan, Mehak Afzal’s team coded in-depth interviews in ATLAS.ti and used Evidano for theme extraction, then reviewed the generated themes themselves — merging duplicates and clarifying the ones that needed further interpretation.

“Evidano was useful in assisting with thematic analysis since it helped in grouping codes and generating themes. It helped organize patterns within the qualitative data and facilitated theme identification. However, AI-assisted analysis should complement rather than replace the researcher’s judgement and interpretation.”

Racha Zeitoun

Doctor of Education, LaFetra College of Education
University of La Verne

For a theory-based dissertation, Racha Zeitoun mapped the Actualized Leadership Profile against transformational, adaptive, connective, and servant leadership through a qualitative comparative analysis of the foundational literature — the kind of document-heavy conceptual work Evidano is built to support.

“I see AI-assisted analysis as a useful support tool rather than a substitute for scholarly judgment. It can help identify patterns, organize data, and generate potential coding suggestions, but the researcher has to determine whether those patterns are actually supported by the theory and the data.”

Chau Pham

Independent Researcher
Vietnam

For a thematic analysis of President Ilham Aliyev’s COP29 speeches, Chau Pham used Evidano to visualize recurring patterns in the language — including the prominence of words such as gas, energy, green, and oil — while the coding, theme development, and interpretation stayed with the researcher.

“Evidano was helpful in identifying rhetorical patterns, a crucial step in the initial research phase. Researchers should keep in mind, however, the importance of original critical analysis and the development of a more nuanced argumentation, which cannot be replaced by any AI tool.”

Published in JAMIA Open

Jorie M. Butler

Behavioral Scientist
University of Utah & VA Salt Lake City

Dr. Butler’s team compared a three-stage ChatGPT workflow for thematic analysis of healthcare interviews against a human-only reference coded by four experienced researchers. The LLM matched most codes and themes in about 5 person-hours instead of 39, but produced broader codes and sometimes framed themes differently.

“At this point, I would say it is easy to see superficial results as better than they are, so I think reviewing the LLM-generated codes in iterations is really key, particularly when there is not a full human-coded set to compare to.”

Published in the Nagoya Journal of Medical Science

Takeshi Kondo

Hospital Lecturer
Nagoya University Hospital, Japan

Dr. Kondo’s team had GPT-4 follow the six steps of thematic analysis on interviews from an earlier medical education study, then compared it with the human-led analysis. Assessors rated ChatGPT highly for traceability, but it stayed close to the surface and let background from the prompt leak into its findings unnoticed.

“AI can amplify the researcher’s own assumptions, and this kind of bias is very hard to see in one’s own work. In our study, it took someone from a different position to see it.”

Published in New Directions for Teaching and Learning

Ruth C. Slotnick

Former Director of Assessment
Bridgewater State University

Dr. Slotnick and Dr. Boeing tested the free 2023 versions of Bard and ChatGPT against their own human-led analysis of program assessment data. The tools were fast and occasionally insightful, but misread charts and frequencies, so the authors built a 10-step guide that keeps verification and interpretation with people.

“AI can help us identify patterns and question our first interpretations. People remain the critical meaning-makers: we need to understand the context, examine what the evidence supports, and take responsibility for what happens next.”

SSRN working paper, 2026

Olivier Toubia

Glaubinger Professor of Business
Columbia Business School

Professor Toubia’s team compared AI-moderated, static and live human interviews with 317 home-improvement decision-makers on a fixed budget. AI moderation recovered customer needs at about half the cost per need of human moderators, but people were more emotionally engaged with humans, and richer interviews did not make digital twins more predictive.

“Insights generated by AI-moderated interviews should be reviewed and audited by humans, and digital twins should be carefully validated before being deployed in a particular use case.”

Published in the Journal of Computer Assisted Learning

Eunhye Shin

Research Fellow
Korea Institute for Curriculum and Evaluation

As a solo teacher-researcher, Shin co-coded 1,287 classroom utterances with ChatGPT-4o. Agreement was moderate (κ = 0.56 against 0.65 between humans) and weakest where context and intent mattered, but asking the model to justify its codes exposed assumptions in her own coding.

“The value was not that the AI was always correct, but that its alternative interpretations helped me verify and reconsider my analysis.”

gLOCAL Evaluation Week 2026 paper

Sergei Gotin

Co-founder & Managing Partner
IMACON Consulting Group, Poland

Gotin’s practitioner framework sorts the AI tools used across one evaluation into four levels by what each is allowed to touch, with anonymisation as the gate before most cloud processing. Every level keeps causal interpretation, conclusions and recommendations with a named human evaluator.

“Be most cautious about the step everyone treats as routine: anonymisation. Replacing names with codes is pseudonymisation, not anonymisation, and in a small context a town or a media outlet can identify a person as surely as a name.”

Published in Anatomical Sciences Education

Michelle D. Lazarus

Professor & Director, Centre for Human Anatomy Education
Monash University

Professor Lazarus and colleagues adapt five qualitative trustworthiness criteria for AI-supported analysis, drawing on a case where a reflexive-language parser grouped student diaries by emotion and only human review could tell whose emotion it was.

“The confusion I felt reviewing the AI categories was a really good reminder of why the human really matters in qualitative research.”