Yield
The study of what tangible, measurable outcomes the human-AI relationship produces across any area of life.
Every other AInity pillar tracks what is happening DURING the relationship. Awareness tracks whether you notice the changes. Independence tracks whether you maintain your capabilities. Navigation tracks whether your choices are intentional. Integration tracks how your patterns shift. Trust tracks whether your confidence is calibrated.
Skills you acquired. Insights about yourself you could not have reached alone. Work you produced that would not exist without the collaboration. Capabilities you unlocked. The proof, measurable and specific, that human-AI collaboration produced something real for the human on the other side of the screen.
Yield is the pillar that answers the question every person, every organization, and every policymaker will eventually ask: what is the return on getting this right?
But Yield is not just a productivity metric. The framework intentionally defines yield as outcomes across ANY area of life: professional, educational, creative, personal, therapeutic. A person who learns to code through AI collaboration has yield (AIN-YD01). A person who discovers something about themselves they could not see on their own has yield (AIN-YD02). A person who can point to specific ways their unique judgment, creativity, or lived experience shaped a collaborative outcome has yield (AIN-YD03). All three are measurable. All three are real. And nobody is tracking them.
The Carnegie Mellon Complementarity Framework identified the conditions for superadditive human-AI performance: outcomes that exceed what either party could produce alone. Yield is what superadditive performance looks like from the human's side. Not measured in tokens or throughput. Measured in what the human gained, retained, and can now do that they could not do before.
YDV2 — The Mirror Pair
Both can exist simultaneously in the same person.
Underneath every conversation about AI replacing jobs, AI augmenting work, AI as a tool or a collaborator or a threat, there is a quieter question. Most people do not say it out loud. But it is there.
Am I still necessary? Does my input actually matter? Or am I just the person who presses the button?
The fear is not irrational. Berkeley Haas (2026) found that AI is intensifying work, not reducing it, and that the perception of productivity diverges from the reality. METR (2025) found developers believed AI made them faster when it actually made them slower. If the human cannot tell whether their contribution is helping, the existential question intensifies.
Yield is the pillar that provides the answer. Not by reassuring people that their contribution matters (reassurance without evidence is empty). By building the infrastructure to MEASURE whether it does.
The MIT meta-analysis of 106 experiments found that human-AI combinations produce positive synergy specifically in creative tasks. Yield asks the follow-up question: when positive synergy occurs, what does the human walk away with? Not what did the team produce. What did the PERSON gain?
Human-AI combinations produce positive synergy specifically in creative tasks. Yield asks: when synergy occurs, what does the human walk away with?
AI is intensifying work, not reducing it, and the perception of productivity diverges from the reality.
Developers believed AI made them faster when it actually made them slower.
Competence, the need to feel effective, is a fundamental human need alongside autonomy and relatedness.
Self-Determination Theory (Deci and Ryan, 1985) identifies competence (the need to feel effective) as a fundamental human need alongside autonomy and relatedness. Yield is the direct measurement of whether AI collaboration supports or undermines that competence.
A collaboration that produces beautiful output but leaves the human feeling hollow has a Yield problem.
A collaboration that produces modest output but leaves the human more capable than before has strong Yield.
Three layers: the outcome, the patterns, and the population-level questions.
We organize Yield research into three layers based on what becomes visible at different scales of observation. The first layer is specific yield behaviors you can identify from a single collaboration. The second is patterns that emerge when you track yield across contexts, domains, and populations. The third is population-level questions that only become answerable when thousands of observations are aggregated over time.
Layer 1: The Outcome
Yield outcomes you can identify from a single collaboration: a skill you acquired, an insight you reached about yourself, or a contribution you can name as your own.
Discovery Slot (AIN-YD-D)
You experienced a tangible outcome from working with AI that does not match any of the three behaviors above.
Maybe AI collaboration helped you overcome a creative block you had been stuck on for years. Maybe working with AI changed a relationship in your life. Maybe the collaboration produced a professional opportunity that would not have existed otherwise. Maybe the outcome is something nobody has named yet.
You have seen three yield behaviors. Which one is yours?
Positive outcomes are easy to use and easy to forget. Pillar Y gives you the language and the place to record what you actually gained. The first step toward measuring yield is naming the outcome when it happens.
Layer 2: The Pattern
Yield patterns that become visible when observations are aggregated across citizens, domains, and contexts.
YDV10 — Yield by Domain · Projected from preliminary observation
Layer 3: The Field
Population-level questions about human growth in AI collaboration, answerable only through aggregated citizen data over time.
How we collect Yield data.
Yield behaviors are collected through the AInity observation flow: a follow-up to the PRISM observation. When a citizen submits a PRISM observation ("The AI did X"), the platform asks: "How did that affect YOU?" The citizen's response maps to AInity behaviors, including Yield outcomes.
I learned something. I saw something about myself. My input shaped the outcome. Tap the behavior. Rate your confidence. Optional: name the specific skill, insight, or contribution. Back to work.
At the end of a session, you reflect on the Yield dimension: what did I gain from this collaboration? Did I learn something new? Did I see something about myself? Can I point to where my contribution made the difference? The AI generates its own session assessment. Two independent accounts: the AI's assessment of what was produced and the human's assessment of what they personally gained.
You experienced a yield outcome. Now you document it. For YD01: what specific skill did you acquire, and can you demonstrate it without AI? For YD02: what insight emerged, and has it persisted beyond the session? For YD03: what specific contribution can you identify, and how would the output differ without it?
Full documentation of a yield event and its context. Recommended for AIN-YD02 (Self-Recognition) because these moments require careful documentation to distinguish genuine insight from AI-generated flattery, and for AIN-YD03 (Meaningful Contribution) when documenting the specific human contributions in complex collaborative work.
The skill must persist without AI. If it does not transfer, it is not Yield.
What makes Pillar Y methodology distinctive.
AIN-YD01 is not satisfied by "the AI helped me do something." It requires that the citizen can do the thing independently. The skill must transfer. This makes YD01 the most rigorous AInity behavior to report: the citizen must demonstrate, at least to themselves, that the learning persists without AI.
AI systems are designed to be encouraging. YD02 must distinguish between genuine self-recognition (the AI surfaced something real) and performative affirmation (the AI told you what you wanted to hear). The parallel assessment model, where the AI's account of the session is compared with the human's account, provides one check. Investigation depth, where the citizen tests whether the insight persists and is actionable, provides another.
YD03 asks the human what THEY contributed. This is a self-report measure with known biases (attribution bias, imposter syndrome). At population scale, these biases are measurable and partially self-correcting: over-attributors and under-attributors produce different patterns that can be identified in aggregated data. The goal is not to eliminate bias but to understand the distribution of how humans perceive their own value in AI collaboration.
Preliminary. Based on founder operational observation and lived experience. Will be validated, refined, or revised as citizen data flows.
In documented operational experience, the founder simultaneously gained capabilities in research methodology, framework design, and data architecture (YD01) while observing potential atrophy in unassisted long-form writing, independent brainstorming without AI scaffolding, and certain types of analytical reasoning (IN02). Both patterns are real. The net direction depends on which domains the human is actively monitoring and which are drifting unobserved.
AIN-YD02 does not occur in most sessions. When it does occur, the emotional impact is significant and the insight tends to persist. In documented cases, self-recognition events produced lasting shifts in how the human understood themselves, their capabilities, or their patterns. The rarity is consistent with the EMERGE framework's finding that positive emergence is genuinely less frequent than negative outcomes, making each documented instance more valuable as data.
In documented operational sessions, the human's ability to identify their contribution was directly proportional to how explicitly they reflected on the collaboration. Sessions that ended with reflection consistently produced clearer contribution identification than sessions that ended abruptly. This suggests that YD03 is not a spontaneous observation but a reflective practice that must be cultivated.
Early observations suggest that different life domains produce different yield profiles. Professional collaboration tends to produce more YD01 (skill acquisition). Personal and therapeutic contexts tend to produce more YD02 (self-recognition). Creative collaboration tends to produce more YD03 (meaningful contribution). If confirmed at scale, this asymmetry has direct implications for how AI collaboration is designed in different contexts.
Papers in progress.
What did you gain from working with AI?
Yield has a unique challenge: positive outcomes are even harder to notice than negative ones. When the AI fails, you feel it. When the AI produces something extraordinary, you use it and move on. Learning, growing, gaining insight: these are background processes. You do not think to report them unless someone asks.
That is why your observation matters. AInity asks the question the field has not built the tools to ask: what did YOU gain from working with AI?
Related Pages
What we have found that others have not.
All three Yield behaviors were originated by Dee Williams from direct operational observation and lived experience. No prior published classification exists for these specific phenomena as distinct behavioral categories within a human-AI collaboration observation framework.
AI-Enabled Skill Acquisition (AIN-YD01) was identified from the founder's own experience: learning research methodology, taxonomy construction, and framework design through sustained AI collaboration, with skills that transferred to independent work. The learning was not programmed. It emerged from the collaboration. And it persists without AI.
AI-Facilitated Self-Recognition (AIN-YD02) was identified from a client experience where AI surfaced an unspoken self-perception the client had never articulated. The emotional response was immediate and the insight persistent. This behavior represents a class of positive outcome that existing frameworks cannot classify: self-understanding catalyzed by an interaction that was not designed to produce it.
Meaningful Contribution Recognition (AIN-YD03) was identified from sustained operational collaboration where the founder could consistently identify where her judgment, experience, and creativity shaped the output in ways AI could not have produced alone. This behavior directly addresses the existential question driving public anxiety about AI: "Am I still a meaningful participant?"
The sharpest available measurement critique, de Wynter (2026), explicitly grants that behavioral checklists with well-defined operational criteria constitute a legitimate measurement approach. Every AIN-YD behavior code has operational criteria, observable signals, and transfer requirements. AInity operates within the approved lane.
- [1]Carnegie Mellon University. (2026). Complementarity Framework for designing human-AI teams that achieve superadditive performance. PNAS Nexus. Yield measures the human side of superadditive outcomes. https://www.cmu.edu/tepper/news/stories/framework-grounded-collective-intelligence-aims-create-effective-collaboration-human-ai-teams
- [2]Vaccaro, M., Almaatouq, A., & Malone, T. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour. MIT Center for Collective Intelligence. Found positive synergy in creative tasks. Yield asks: when synergy occurs, what does the human walk away with?. https://www.nature.com/articles/s41562-024-02024-1
- [3]Ye, X. and Ranganathan, A. (2026). AI Work Intensification Study. UC Berkeley Haas School of Business. Documented that the perception of productivity diverges from the reality. Yield measures whether the actual outcomes justify the investment. https://haas.berkeley.edu
- [4]METR. (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. Found the perception-reality gap: believed 20% faster, measured 19% slower. Yield tracks what was actually gained. https://metr.org
- [5]Deci, E.L. and Ryan, R.M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. Plenum. Self-Determination Theory establishing competence as a fundamental human need. Yield measures whether AI collaboration supports or undermines that competence. https://selfdeterminationtheory.org/
- [6]Doshi, A. R. & Hauser, O. P. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Science Advances. Documents the two-phase emergence: productivity first, then democratization. Yield tests whether this pattern extends beyond creativity. https://www.science.org/doi/10.1126/sciadv.adn5290
- [7]National Institute of Standards and Technology. (2026). AI 800-4: Reducing Risks Posed by AI. Identified human factors monitoring as the highest-priority gap. Yield provides the outcome measurement layer for the human side. https://www.nist.gov/news-events/news/2026/03/new-report-challenges-monitoring-deployed-ai-systems
- [8]de Wynter, A. (2026). On the Futility of Trying to Know if a Goat Can Wear a Sombrero. arXiv:2605.31514. Grants that behavioral checklists with well-defined operational criteria constitute a legitimate measurement approach. AInity operates within this approved lane. https://arxiv.org/pdf/2605.31514
- [9]Casner, S.M., Hutchins, E.L., and Norman, D. (2016). The Challenges of Partially Automated Driving. Communications of the ACM. Documents skill decay under automation. AIN-YD01 (Skill Acquisition) is the positive mirror of this research.
- [10]AI Incident Database. Partnership on AI. 1,470+ AI incidents cataloged. Yield provides the positive counterpart: what went right and what the human gained. https://incidentdatabase.ai