Epistemic Agency: AI Qualitative Analysis of Self-Tracking

Epistemic agency in self-tracking is a framework for deciding whether individuals merely provide measurements or actually control knowledge production. Researchers and UX teams evaluating citizen science for health face a measurement problem: how to classify and synthesize studies that range from device-driven data collection to fully self-directed "personal science" inquiries. The primary keyword for this post is epistemic agency in self-tracking, and this post explains the analytic counts, the four-phase framework used in the source study, and how AI-enabled qualitative research can operationalize those distinctions for robust synthesis and actionable insight.

Key Takeaways

According to the Frontiers in Public Health article published on 18 August 2026, an agency-based, four-phase framework (Questioning, Observing, Reasoning, Discovering) distinguishes low-, mixed-, and high-agency forms of self-tracking and shows that personal science sits at the high-agency end (Frontiers in Public Health).

In April–May 2025 the authors searched Web of Science and PubMed for “quantified self” and retrieved 721 unique records, of which 241 full papers were retained for analysis.
Of the 241 retained papers, 46 used self-tracking as the data-collection method and were classified by agency level: 17 low-agency, 10 mixed-agency, and 19 high-agency (classification reported in the August 18, 2026 article).
The framework turns on one key question: “who owns the question”; when the individual owns the question the practice maps to personal science and high epistemic agency (Frontiers in Public Health, 18 Aug 2026).

What happened and how the framework works

Answer: The Frontiers in Public Health review (published 18 August 2026) applied a four-phase personal science framework to classify how much control individuals had over self-tracking studies.

According to the Frontiers in Public Health article, the authors adapted an existing five-phase model into four phases (Questioning, Observing, Reasoning, Discovering) to match observed workflows and then applied three agency levels (low, mixed, high) to 46 empirical self-tracking studies identified from a larger corpus.

Method details: the review searched Web of Science and PubMed in April–May 2025, retrieved 721 records, screened 618 abstracts, resolved 13% reviewer conflicts (n = 80) by group discussion, and retained 241 full papers for final analysis; 46 of those used self-tracking for data collection and were eligible for agency coding.

Constraints and conservative coding: the Frontiers authors assigned the lower agency level when evidence was ambiguous, and they treated co-authorship by the self-tracker as a practical indicator that Discovering was self-directed.

Findings Snapshot

Date	Metric	Value	Implication
April–May 2025	Records retrieved (search term: "quantified self")	721	Large heterogeneous literature requiring synthesis; justifies an agency-based lens
April–May 2025	Full papers retained after screening	241	Corpus includes both empirical and conceptual work; only a subset contains self-tracking data
August 18, 2026	Papers using self-tracking as data-collection method (analyzed for agency)	46	Allows direct assessment of who performs each inquiry phase
August 18, 2026	Agency breakdown among the 46 studies	17 low, 10 mixed, 19 high	Almost half of empirical self-tracking studies display high individual agency

Implications for researchers and UX teams

Answer: The agency distinctions change how teams design, evaluate, and synthesize self-tracking research.

Practical research implications: according to the Frontiers in Public Health article (18 Aug 2026), low-agency studies typically treat participants as sensor platforms and emphasize data sharing, while high-agency studies emphasize sharing methods, learnings, and co-authorship, which affects how qualitative themes should be interpreted and weighted in synthesis.

Design and product implications: UX teams should distinguish instrumentation (device accuracy) from epistemic design (who defines questions and interprets results), because devices that encourage user-defined Questioning and Reasoning increase epistemic agency and produce different qualitative data.

Evidence synthesis implication: reviewers should code for agency level (Questioning, Observing, Reasoning, Discovering) as categorical metadata, because the Frontiers in Public Health review shows that collapsing all “self-tracking” studies obscures whether the study produced self-directed knowledge or merely measurement.

How Evidano helps AI-enable qualitative synthesis

Problem: Heterogeneous corpus, slow synthesis → Solution: Thematic + agency coding

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

Evidano feature mapping: apply automated document ingestion to load the 241-paper corpus, then run custom theme extraction and the four-phase agency classifier to label Questioning, Observing, Reasoning, and Discovering at scale.

Operational benefit: the Frontiers review needed manual dual-author coding and conflict resolution; Evidano reduces that time by producing reproducible, auditable agency labels and by surfacing candidate quotations for human verification.

Problem: Fragmented methods and missing contextual data → Solution: Cross-segment and co-occurrence analysis

Evidano can perform cross-segment comparisons (for example, device type versus agency level) and co-occurrence network visualizations so researchers can see whether consumer wearables correlate with lower or higher agency across studies.

Contextual workflows: teams can export labeled excerpts and visualizations for inclusion in systematic reviews or UX decision decks, or use Evidano's features page to map tools to methods.

Problem: Multi-format source material (figures, transcripts, logs) → Solution: Unified ingestion and searchable corpus

Evidano ingests PDFs, transcripts, and spreadsheets, enabling reviewers to align qualitative descriptions (for example, "sharing of methods") with numeric metadata (agency counts) from the Frontiers dataset.

For studies that include speech or interview audio, Evidano's speech-to-text can transcribe with custom dictionaries, preserving terms like "quantified self" and "personal science" for accurate theme extraction.

Ethics and provenance note

AI-enabled qualitative analysis should preserve provenance and human oversight: the Frontiers article highlights positionality and conservative coding rules, and Evidano supports audit logs and human-in-the-loop validation so agency claims remain defensible.

FAQ: epistemic agency in self-tracking

What is epistemic agency in self-tracking?

Answer: Epistemic agency in self-tracking measures whether an individual controls the knowledge-making cycle rather than merely supplying data.

Supporting detail: The Frontiers in Public Health review (18 Aug 2026) defines epistemic agency as the capacity to pose questions, gather and interpret evidence, reach conclusions, and be recognized as a credible knower; the review operationalizes this across four phases: Questioning, Observing, Reasoning, Discovering.

How did the Frontiers review classify studies by agency?

Answer: The review used a four-phase personal science framework and three agency levels to code papers based on who performed each phase of inquiry.

Supporting detail: According to the Frontiers article, reviewers checked whether Questioning, Reasoning, and Discovering were performed by the self-tracker (high agency) or by researchers (low agency); co-authorship by a self-tracker was used as a practical indicator of self-directed Discovering.

Which quantitative signals indicate personal science in the literature?

Answer: Numbers reported in the review indicate scope and prevalence: 721 records retrieved (April–May 2025), 241 full papers retained, 46 empirical self-tracking studies, and an agency split of 17 low, 10 mixed, 19 high (Frontiers in Public Health, 18 Aug 2026).

Supporting detail: These counts show that nearly 41% of the 46 empirical self-tracking studies were high-agency, suggesting personal science is a substantial, visible strand within the quantified self literature.

Can AI reliably classify agency in self-tracking studies?

Answer: AI can accelerate agency classification but should operate with human validation and transparent coding rules.

Supporting detail: The Frontiers authors used independent dual-author coding with conservative rules; AI-enabled pipelines like Evidano can reproduce those rules at scale, surface conflicts, and reduce human workload while preserving the need for human adjudication.

Conclusion & Next Steps

The Frontiers in Public Health review (published 18 August 2026) demonstrates that an agency-based framework meaningfully separates mere measurement from self-directed personal science, and it shows concrete counts: 721 records retrieved in April–May 2025, 241 full papers retained, 46 empirical self-tracking studies, and an agency split of 17 low, 10 mixed, 19 high.

For teams conducting qualitative synthesis or building citizen science programs, the practical next step is to code agency across your corpus, prioritize studies that include Questioning and Discovering by participants, and surface method sharing and co-authorship as signals of high epistemic agency.

To accelerate that work with AI-enabled qualitative analysis and reproducible agency labeling, consider using Evidano's thematic and cross-segment tools and transcription capabilities; see the Evidano features page for details.

Get started: Try Evidano for free.

Quote (conceptual anchor): Catherine Elgin, quoted in the Frontiers review, wrote, "Epistemic agents should think of themselves as, and act as, legislating members of a realm of epistemic ends, " and Heyen called personal science the production of "verified and practical self-knowledge."

Topics

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

Keep reading

Browse all articles