How to integrate AI into your research flow with integrity.
The HEARTS Framework is a six-part method for deciding whether an AI-assisted research workflow is actually better — not just faster. It’s how we, as UX researchers, incorporate AI with intention and integrity.
Six lenses. Six ideals.
HEARTS is an acronym for six design principles you use to design and evaluate your AI-assisted workflows. Each letter is a lens — asks you to reflect and ensure your workflow is truly better, not just faster.
Human-led & centered
The researcher is the pilot, not the passenger. AI is a tool directed by and for humans. We prioritize the needs, context, and well-being of everyone in the research process — practitioners, participants, and partners.
Where does the human stay in control? If the answer is “nowhere,” you’ve built an automation, not an AI-assisted workflow.
Experience-focused
We are accountable for the quality of the experience. Every interaction with AI and its outputs — by a researcher, a participant, or a stakeholder — must be intuitive, respectful, and positive.
Does this improve the experience for everyone it touches, or just speed things up for you?
Amplification, not automation
AI amplifies researchers’ superpowers — you just have to earn it. Delegate the repetitive and unnecessary; protect the brainpower for critical thinking, strategic synthesis, and building empathy.
Is AI freeing the researcher to do something more valuable, or just doing the valuable part for them?
Rigorous & responsible
Maintain integrity, minimize bias, maximize confidence. We proactively mitigate bias and maximize accuracy in every tool and process. We don’t “check the box” on ethics; we build our practice around it.
How are you validating accuracy and checking for bias — your own and the model’s?
Trustworthy & transparent
Transparency is the bedrock of trust; traceability is the bridge. From collection to analysis, we provide a clear path for verification and prove the lineage of our insights. We disclose by default.
If someone audited this process tomorrow, could they follow the trail?
Safe, secure & sustainable
Protect the people, the data, the practice, and the planet. Safe — the well-being of everyone involved. Secure — zero-trust for their data, no PII in public models. Sustainable — workflows your team can maintain.
Are you protecting people’s safety, securing their data, and building something that lasts?
The HEARTS Scorecard
What's the integrity of your AI-assisted research workflows?
Score one AI-assisted workflow across the six HEARTS lenses. You'll get a per-lens profile, any floor conditions to fix first, and the one lens worth focusing on next.
How this works & what it can't tell you
- You'll answer 15 questions, one at a time, across the six lenses — each on a 1–5 frequency scale (1 = Never, 5 = Always). Pick "Not sure" if you can't see it. On most questions it's left out of the math. On the four floor questions, not knowing counts as the gap.
- Each lens is the average of its questions. Your result leads with the six-lens profile, not a single score.
- Floor conditions come first. Four floor questions cover risks a strong score elsewhere can't buy back: unchecked outputs, no human review, undisclosed AI, and personal data reaching AI tools unprotected. If one is live, you'll see it before anything else.
- The 2×2 is a secondary view. Each axis — Rigor (H, A, R) and Responsibility (E, T, S) — takes your lowest lens on that side, not the average, so one strong lens can't hide a weak one. It places you as an Improviser, Maverick, Idealist, or Steward. Directional, not precise.
- Questions ask how often each practice actually happens — easier to answer honestly than agree/disagree.
- At the end, a few optional questions about your team help build anonymous benchmarks — used in aggregate only, and linked to you only if you ask for your results by email.
- Status: beta. HEARTS scoring is pre-validation — evidence-informed, not yet validated.
- Limitation: a self-assessment from one person's view — a structured reflection to surface blind spots, not a formal audit or a measure of research quality. For a team-level read, the HEARTS Audit walks one of your team's workflows through all six lenses. Book a free call to run it together.
Your answers are saved without your name or email to build anonymous AIxUXR benchmarks. Untick below to keep them out of the benchmarks. Privacy policy
A little about you and your team
These help build anonymous AIxUXR benchmarks, so you'll be able to compare your results to similar teams down the line. Used in aggregate only — skip any you like. Saved only if you left benchmarks on at the start, and linked to you only if you later ask for your results by email.
A strong score elsewhere can't make up for these. Under the HEARTS rubric, a workflow with a live floor condition isn't ready to trust until it's fixed.
Want this as a takeaway?
Get your profile, any floor conditions, and your focus lens by email.
What happens to your answers
If benchmarks are on, your answers and optional team details are saved to a private Google Sheet without your name or email. If you ask for your results by email, your email is saved with your results — and, if benchmarks are on, with your answers and team details — and used to send them. Resources are sent only if you tick the box above; reply "unsubscribe" anytime. Booking a call passes a short result summary to the booking form. To see or delete your data, email kaleb@heykaleb.com. Full details are in my privacy policy.
Now what?
A checklist, plus the HEARTS Pulse Check prompt that reviews your real work. Free when you join the newsletter. You can unsubscribe anytime.
Explore the full HEARTS frameworkKnow a team adopting AI in research? Save your card and challenge them to find their quadrant.
A score of 2 or below on R, T, or S means you redesign before you collect data—not after. Those three are where the damage is hardest to undo. Re-run the scorecard after launch to see whether the rigor held.
The questions above are the HEARTS Pulse Check: one question per lens, for a fast check. For a full audit, HEARTS In-Depth breaks each lens into six questions, 36 in all. Each is scored from 0 to 5, then averaged into the lens score. It's the rubric I used to score my own AI pipeline in The ResearchOps Review. The In-Depth rubric is still being tested, so I walk teams through it live.
A free 30-minute call. Bring the workflow, not the data.
HEARTS aligns with the standards — and adds two they miss.
HEARTS translates the Responsible-AI consensus and re-centers it for researcher workflows. It aligns with the major frameworks on most dimensions, and adds two the compliance frameworks miss: Experience and Amplification.
| HEARTS | OECD (2024) | NIST AI RMF | EU AI Act | Academic Consensus |
|---|---|---|---|---|
| Human-Led | Human-centred values | Human role in accountability | Human oversight (Art. 14) | Human oversight & control |
| Experience-Focused | Well-being (partial) | — | — | UX-specific |
| Amplification | Inclusive growth (partial) | — | — | Human–AI complementarity |
| Rigorous & Responsible | Robustness; fairness | Valid & Reliable; Fair | Accuracy, robustness | Fairness & non-discrimination |
| Trustworthy & Transparent | Transparency; accountability | Accountable & Transparent | Transparency (Art. 50) | Transparency; accountability |
| Safe, Secure & Sustainable | Security & safety | Safe; Secure & Resilient | Safety; data governance | Privacy, security & well-being |
Built from experience, not a blank page.
While leading Responsible-AI research at Instacart, I found that OECD, NIST, the EU AI Act, and the academic literature all circle the same core — keep humans in control, be fair, be transparent, protect data, don’t harm people or the planet. But that consensus was written for governments and engineering orgs.
Not for the researcher staring at a transcript, wondering if the AI summary is trustworthy.
HEARTS translates it for the researcher doing the work — and adds the two things the compliance frameworks miss. First announced at the AI Club webinar on Sept 26, 2025, and pressure-tested since in enterprise workshops, the AI in UXR 101 course, and the AIxUXR community.
Questions.
What is the HEARTS Framework?
A six-part framework for evaluating whether an AI-assisted research workflow is rigorous, ethical, and trustworthy — not just faster.
Who created it?
Kaleb Loosbrock, a Staff/Principal-level UX researcher and AIxUXR consultant — based on research and experience integrating AI into his workflows at Instacart, and refined with the AIxUXR community.
How is it different from NIST or the EU AI Act?
HEARTS aligns with them, but it’s built for the working researcher — and adds two dimensions they don’t name: Experience and Amplification.
Does it replace Responsible AI standards?
No. It operationalizes the Responsible-AI consensus (OECD, NIST, EU AI Act) for research practice; it complements those standards.