What the study analysed
The mixed-methods study examined which leadership credentials the public values when evaluating police chief candidates. A discrete-choice conjoint experiment with 303 respondents was fielded in April 2026, followed roughly two weeks later by an open-ended survey asking respondents to explain how they distinguished between pairs of applicant qualifications.
The follow-up produced 224 complete responses across two questions, each at least 100 characters. Thematic analysis of the professional standards question produced five themes; the evidence-based policing question produced four.
Manual coding first, AI as the cross-check
The first author read every response, generated codes inductively, and built, refined, and defined the themes by hand. Only then was the qualitative data run through Evidano, to triangulate those findings and to help identify oversights, alternative frames, and personal biases.
Eight of the nine human-coded themes were supported by the AI-assisted analysis with only minor differences in wording, and the ninth was folded into two others as a countervailing perspective. Evidano also surfaced two candidate themes the human coding had not highlighted, one of which the team incorporated as a caveat and one it rejected because the underlying responses did not cohere. All coding and thematic interpretations were reviewed and validated by the research team.
The paper explains why the tool was chosen: five recent peer-reviewed studies had demonstrated its ability to support accurate, low-bias analysis when used with human oversight.
How Evidano fit the workflow
A human-led thematic analysis, cross-checked rather than replaced.
- 1Open-ended responses collected
224 complete responses to two short-answer questions, following a 303-respondent conjoint survey.
- 2Inductive coding by the first author
Codes were generated from the data, then renamed, combined, or split as new meanings emerged.
- 3Themes generated, refined, and defined
Five themes on professional standards, four on evidence-based policing.
- 4Evidano used to triangulate
A second pass to surface oversights, alternative frames, and personal biases.
- 5Research team reviewed and validated
All coding and thematic interpretations were checked by the team, which kept the final call on every theme.
“we analyzed our qualitative data using the online AI-powered qualitative analysis software Evidano (formerly known as AIlyze) to triangulate the findings from our thematic analysis and to "help identify oversights, alternative frames, and personal biases"”
“Evidano was selected because recent qualitative studies […] have demonstrated its ability to support accurate, low-bias qualitative analysis, particularly when used with human oversight.”
“All four themes from the human-coded analysis […] were supported by the AI-assisted analysis with only minor differences in wording.”
“Accordingly, all coding and thematic interpretations were reviewed and validated by the research team.”
What a triangulation pass adds
One analyst coding hundreds of free-text responses carries their own framing into every theme, with nothing to test it against.
An independent pass over the same dataset that surfaces oversights and alternative frames before the findings are published.
The study names Evidano, cites it, and explains why it was selected.
Themes are identified across the entire dataset rather than response by response.
User data is confidential and never used to train models.
Every coding and thematic interpretation was reviewed and validated by the research team.