Epistemic Agency: AI Qualitative Analysis of Quantified Self

This post explains how to assess "epistemic agency in self-tracking" for qualitative researchers and product teams, using the August 18, 2026 Frontiers review as the empirical anchor. According to the Frontiersin.org review, applying a four-phase personal science framework (Questioning, Observing, Reasoning, Discovering) lets researchers separate low-agency measurement from high-agency personal science, and the review quantified that separation across 241 full papers and 46 self-tracking studies. If your team runs interviews, open-ended survey responses, or participant-generated logs, AI-enabled qualitative analysis can scale identification of who "owns the question" and surface patterns across N-of-1 cases.

Key Takeaways

According to the Frontiersin.org review (published 18 Aug 2026), an agency-based framework distinguishes "measurement" from self-directed "personal science", and the authors conclude that "The framework can meaningfully distinguish merely measurement from self-directed knowledge production." Frontiersin.org.

A Web of Science and PubMed search in April–May 2025 returned 721 records and yielded 241 full papers for analysis, as reported by the Frontiersin.org review.
Among the 241 papers, the Frontiersin.org analysis found 46 studies that used self-tracking as the data collection method, of which 17 were low-agency, 10 were mixed-agency, and 19 were high-agency (Frontiersin.org, 18 Aug 2026).
The Frontiersin.org authors emphasize the core distinction as "who owns the question" and invoke the philosophical standard that "Epistemic agents should think of themselves as, and act as, legislating members of a realm of epistemic ends" (Catherine Elgin), a quotation used in the review.

What happened: epistemic agency in self-tracking according to the Frontiers review

Answer: The Frontiersin.org conceptual review applied a four-phase personal science framework to classify levels of individual agency in self-tracking studies and quantified the distribution across the literature.

According to the Frontiersin.org article published on 18 Aug 2026, the authors used Jaakkola's theory-synthesis approach and an adapted four-phase framework (Questioning, Observing, Reasoning, Discovering) to evaluate epistemic agency across studies in health literature.

According to the Frontiersin.org methods section, the authors searched Web of Science and PubMed in April–May 2025 for the term "quantified self", found 721 unique records, screened down to 241 full articles, and then identified 46 studies where self-tracking was the data-collection method for detailed agency coding.

According to the Frontiersin.org results, the 46 self-tracking studies split into 17 low-agency (participants only collected data), 10 mixed-agency (participants did more than collect data but not the full cycle), and 19 high-agency (participants performed Questioning, Observing, Reasoning, and Discovering).

According to the Frontiersin.org discussion, the review's conservative coding rules meant that when a phase could reasonably be attributed to either participant or researcher, the lower agency level was recorded to avoid overstating participant control.

Findings Snapshot

Date	Metric	Value	Implication
April–May 2025	Records retrieved with "quantified self"	721	Large initial corpus, justifying targeted screening (Frontiersin.org)
April–May 2025	Full papers retained after screening	241	Corpus for concept-level synthesis (Frontiersin.org)
–	Studies using self-tracking as data-collection	46	Subset eligible for agency coding (Frontiersin.org)
18 Aug 2026	Agency distribution (low/mixed/high)	17 / 10 / 19	Roughly 41% low, 22% mixed, 41% high agency in the 46 self-tracking studies (Frontiersin.org)
Screening stage	Reviewer disagreement during abstract screening	13% (n = 80)	Conflicts resolved by discussion among authors (Frontiersin.org)

Implications for qualitative researchers and UX teams

Answer: The Frontiersin.org framework shows qualitative teams must register who frames the research question if they want to claim participant epistemic agency.

According to the Frontiersin.org review (18 Aug 2026), low-agency studies typically treat participants as "sensor platforms" where investigators define questions and perform Reasoning and Discovering, which reduces participant epistemic standing.

According to the Frontiersin.org analysis, mixed-agency articles reveal hybrid practices where devices and platforms co-produce interpretation, and high-agency articles show participants conducting N-of-1 cycles, often co-authoring publications.

According to the Frontiersin.org discussion, qualitative researchers and UX teams who want to support personal science should design studies that enable participants to own Questioning, provide tools for participant-led Reasoning (visualization and simple analytics), and create pathways for participant-led Discovering and dissemination.

How Evidano helps teams assess epistemic agency in self-tracking

Problem: Identifying who owns the question across hundreds of participant narratives

Answer: Teams struggle to scale manual coding to determine whether participants formulated questions, performed analysis, or only supplied data.

According to the Frontiersin.org review, distinguishing Questioning from Observing is essential to locate epistemic agency, but manual review of 241 papers was labor intensive (Frontiersin.org).

Solution: Thematic + cross-segment analysis

Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents.

Evidano maps the Frontiersin.org framework into operational codes: Questioning, Observing, Reasoning, Discovering, then runs thematic, frequency, and cross-segment analyses to show who performs each phase across transcripts or logs.

Teams can use Evidano to convert long participant diaries, meetup transcripts, or forum posts into structured evidence showing when participants "own the question" and when they act as data sources.

Feature fit

Problem: Manual transcription and inconsistent terminology obscure participant methods; Solution: use Evidano's speech-to-text with custom dictionaries and PII redaction to produce clean transcripts that preserve method language.

Problem: Multimodal inputs and multilingual posts are hard to analyze; Solution: use Evidano's translation and ingestion pipeline to normalize data for cross-case coding.

Problem: Teams need to present findings to clinicians and designers; Solution: Evidano generates visualizations (co-occurrence networks, hierarchical code trees) and exportable reports that map participant agency levels to product or study decisions.

For more on product capabilities see the Evidano features page.

Practical workflow

Step 1: Ingest interviews, survey open-ends, device logs, and forum posts into Evidano.

Step 2: Apply a custom codebook reflecting the Frontiersin.org four-phase framework and run AI-assisted coding to label Questioning, Observing, Reasoning, Discovering.

Step 3: Use cross-segment analysis to quantify agency by cohort, date, or device type and export a concise evidence package for stakeholders.

FAQ: epistemic agency in self-tracking

What is epistemic agency in self-tracking?

Answer: Epistemic agency means an individual exercises control over knowledge production rather than merely supplying data, according to the Frontiersin.org review (18 Aug 2026).

According to the Frontiersin.org article, the review operationalizes epistemic agency as whether a person performs Questioning, Observing, Reasoning, and Discovering in their own self-tracking practice.

How did the Frontiersin.org review measure levels of agency?

Answer: The authors coded papers against a four-phase framework and assigned three agency levels (low, mixed, high), as described in the Frontiersin.org methods (18 Aug 2026).

According to the Frontiersin.org methods, Observing was present in all included self-tracking studies by definition, while Questioning, Reasoning, and Discovering were assessed for participant vs researcher ownership.

What are practical signs of high-agency personal science in transcripts?

Answer: High-agency transcripts show participant-formulated questions, iterative method changes, participant-led analysis language, and self-initiated dissemination, as reported by Frontiersin.org.

According to the Frontiersin.org findings, a practical signal of high agency was when the self-tracker was a co-author of the resulting publication, indicating participant-led Discovering.

Can AI help detect agency automatically?

Answer: Yes, AI-enabled qualitative analysis can scale detection of agency markers across large corpora by combining codebooks with semantic search and pattern extraction.

According to the Frontiersin.org review, large-scale syntheses require consistent operationalization of phases; Evidano's AI coding and cross-segment analytics implement that approach to find who frames questions and who interprets data across cases.

Conclusion & Next Steps

The Frontiersin.org review (published 18 Aug 2026) shows that an agency-based, four-phase framework meaningfully separates measurement from participant-led personal science, and the review quantified that 46 self-tracking studies included all or part of that spectrum.

Qualitative teams and UX researchers can use this framework to redesign studies so participants can own Questioning, Reasoning, and Discovering rather than only Observing, an approach the Frontiersin.org authors recommend.

Evidano operationalizes that recommendation with AI-assisted coding, transcript-quality features, and cross-segment visualizations to show at scale who "owns the question" across interviews and logs.

Ready to run an agency analysis on your corpus? Try Evidano for free.

Topics

epistemic agency in self-tracking
personal science qualitative analysis
quantified self research
AI-enabled qualitative research

Keep reading

Browse all articles