COMMONS FOR PARTNERS: METHODS

How we think, build, vet, design, test, and revise. If you're evaluating this work, you deserve to know how it's made.

steamHouse makes large claims. We say we've synthesized 1,100+ sources across 48 research domains into a coherent developmental framework. We say this framework can be used by any mentor in any team or project context. We say it addresses the central educational challenge of the coming century.

Those are claims worth scrutinizing. This page explains the methods behind them — not just what we've built, but how we build, how we vet, and how we plan to test whether any of it works.

How We Think

Procedural, Not Substantive

steamHouse's epistemological position is that we teach the process of seeking truth rather than endorsing specific conclusions. On contested questions — where smart, informed people disagree — we present the strongest versions of competing positions and teach young people to evaluate evidence, recognize uncertainty, and revise their thinking.

This isn't relativism. We hold that some claims are better supported than others, that evidence matters, and that intellectual honesty requires acknowledging what you don't know. But we distinguish between procedural commitments (how to think) and substantive positions (what to conclude). We're confident about the procedures. We're appropriately humble about conclusions — especially in domains where the research itself is contested.

This shapes every curriculum decision. When our bias compendiums cover topics where experts disagree, we present the disagreement rather than resolving it. When our developmental markers describe capacities, we describe observable behaviors rather than inner states we can't verify. The framework is designed to be robust across different worldviews, not dependent on any particular one.

→ Full epistemological position statement

How We Build

The Seven-Phase Composition Pipeline

steamHouse's intellectual architecture lives in a library of books, compendiums, and curriculum materials. These are produced through a structured human-AI collaboration using Claude (Anthropic) as the primary composition tool.

AI-assisted composition at this scale is new. Best practices are still emerging — for us and for everyone. We describe our process with complete transparency because researchers, funders, and partners deserve to know what they're evaluating, and because we expect our methods to continue evolving as the tools mature and as we learn more about what produces genuinely reliable output. Near-term priorities include ongoing evaluation of both the process itself and the quality of what it produces — not as a one-time audit but as a standing commitment.

Phase 1: Source Curation — Human judgment only. No AI involvement. Our research bibliography currently spans approximately 1,100 sources across 48 domain categories. Sources are selected for peer-reviewed or peer-respected status, author credibility, relevance to the steamHouse developmental framework, and balance across perspectives within contested domains.

Phase 2: Deep Dive Reference Summaries — Claude generates structured analyses of each major source, drawing on its training data. These summaries are strong for well-documented works and degrade for obscure or very recent publications. The human reviews and corrects.

Phase 3: Compendium Creation — The most labor-intensive phase. Multiple sources within a domain get synthesized into thematic documents — not annotated bibliographies but genuine syntheses that identify convergences, tensions, and patterns across sources. 48 domain compendiums exist, totaling approximately 1.6 megabytes of synthesized research content. Human determines structure and synthesis judgments; Claude drafts and refines through multiple passes.

Phase 4: Grounding the AI — Compendium documents are loaded into Claude's project knowledge environment, giving it access to steamHouse's curated research base rather than undifferentiated training data during subsequent work.

Phase 5: Outline Planning — Human designs scope (what's included and excluded), sequence (what order material is encountered), and angle (voice, frame, relationship with reader). Claude consults and pressure-tests. Final decisions are human.

Phase 6: AI Draft Generation — Claude produces structured draft content, chapter by chapter, pulling from the compendium base and following human-approved outlines. This is where AI's role is most visible as drafting.

Phase 7: Human-AI Iterative Editing — Human review, substantive revision, dialogue about specific passages, and multiple improvement cycles.

Current status, stated plainly: Phase 7 has not been completed for any steamHouse book. What exists are strong AI-generated drafts shaped by human outlines and grounded in human-curated research. The research foundation is solid. The book drafts are working documents, not finished manuscripts.

The most accurate description of our materials: human-directed, AI-drafted, human-edited — with editing still underway across all titles.

→ Full AI Composition Methodology statement

How We Vet

Research Standards and Synthesis Methods

steamHouse synthesizes across domains that don't normally talk to each other — cognitive science, evolutionary psychology, mentoring research, systems thinking, character development, adolescent neuroscience, and more. This creates a synthesis challenge: how do you integrate findings from fields with different methodological standards, different evidence norms, and different vocabularies?

Our approach:

Source hierarchy. Peer-reviewed research is primary. Well-grounded practitioner literature and rigorous popular science that synthesizes research meaningfully are included when they add genuine insight. We don't weight all sources equally — a meta-analysis carries more weight than a single study, and a replicated finding carries more weight than a novel one.

Convergence over citation. When researchers in different traditions, using different methods, arrive at structurally similar conclusions, we treat that convergence as stronger evidence than any single study. Our "Independent Convergence" analysis documents ten external frameworks that arrive at conclusions consistent with steamHouse's without knowledge of our work. This triangulation approach is how we compensate for not having our own outcome data yet.

Honest handling of disagreement. Where experts disagree, we acknowledge the disagreement rather than cherry-picking the side that supports our framework. The compendiums explicitly mark tensions and unresolved questions. We'd rather be honest about uncertainty than appear more certain than the evidence warrants.

Domain balance. We actively seek sources that challenge or complicate our framework, not just ones that confirm it. The research bibliography includes critics of metacognitive interventions, skeptics of character education, and researchers who question synthesis-based approaches. We believe engaging with the strongest counter-arguments makes our position stronger, not weaker.

→ Research Foundations → Research Bibliography (500+ sources)

How We Design

Think Big, Then Be Real, Then Act

This is a slogan on the home page. It's also a literal description of how steamHouse was built — and a disposition toward the work that shapes every major decision.

Think Big applied reflexively. steamHouse teaches conscious thinking — thinking that deliberately considers more context than you would by default. We'd be hypocrites not to apply that to our own design process. So we did. Before building anything, we asked the widest possible question: What would a complete developmental ecosystem actually require? Not what's fundable. Not what's already being done. What's right — if you were designing from scratch, knowing what research across fifteen domains tells us about how humans actually develop?

That question produced something unexpected. The curriculum didn't get bigger. It got smaller. When you push scope to its limit — "what would every human benefit from learning?" — the answer converges on remarkably few things: four principles, three capacities, present in every domain under different names. The widest possible scope produced the tightest possible focus.

This is what we call the Scope-Focus Paradox, and it's both a design discovery and a core curriculum insight. The breadth was the method. The focus was the discovery.

The logic is similar to what the standards movement attempted in education: consolidate diverse curricula into fewer shared objectives, and you create economies of scale — better resources, better lessons, useful for everyone. steamHouse applies that same consolidation logic not to academic content but to human development itself. Fewer ideas, and better developed. Through-lines that run across every dimension of development, not replacements for those dimensions.

This is why the project looks ambitious to the point of implausibility. The breadth is deliberate — every component exists because the system requires it, not because we were chasing scope. We mapped the whole system first — how curriculum connects to assessment, how assessment connects to credentialing, how credentialing connects to community, how community connects to narrative, how narrative creates the desire that drives people back to curriculum. We designed the relationships before building the components.

Be Real means acknowledging honestly what's built, what's in progress, and what's aspirational. The Workshop section of this site exists precisely for this reason — to show the complete vision alongside the honest status of each piece. We don't present drafts as finished products. We don't describe proposed programs as operational. The Level Set statement on the Workshop page is blunt about where things stand.

Think Big drew the map. Be Real is the patient, honest work of filling it in.

Act means doing what you can to get what you want — and refusing to let what you can't do get in the way of what you can do. steamHouse doesn't have institutional backing, a research team, or a large budget. It built anyway. Not everything. Not perfectly. But real activities with real families in a real community. The Fairmount steamHouse Club in Golden, Colorado has been operating since 2018 — 110+ families, seasonal events, robotics teams, camps. The design isn't theoretical. It's being tested in practice, imperfectly but genuinely, by people who decided that waiting for perfect conditions was its own form of giving up.

→ The Widest Scope, the Tightest Focus (essay) → The Breadth Is the Point (design reference)

How We Test

Measurement Approach

steamHouse does not yet have outcome data from controlled evaluation. We're transparent about this because it's the single most important thing a serious evaluator needs to know.

Here's what we do have and what we're building toward:

What exists now:

A Pilot Measurement Toolkit with standardized instruments for data collection during pilot implementations — participant intake, mentor fidelity logs, monthly reflections, development marker self-assessments, and exit surveys

Three data collection packages scaled to partner capacity: Minimal (15 minutes total participant burden), Standard (60-90 minutes over full pilot), and Research-Ready (90-120 minutes with mentor ratings and parent surveys)

The 58 Development Markers functioning as standards-aligned assessment constructs — observable behaviors organized into Stars (character capacities), Lenses (ways of seeing), and Keys (demonstrated skills)

Eleven interactive tools that generate individual assessment data (Thinking Bias Profiler, Authorship Assessment, etc.) — not yet validated as psychometric instruments, but designed to produce comparable data across users

What we need:

Research partners to design and conduct rigorous evaluation

Pilot sites willing to implement the Standard or Research-Ready data packages

Formal validation of the Development Markers as assessment constructs

Psychometric validation and norming studies as scale increases

The honest framing: We have design validity (the framework is consistent with established research) and independent convergence (ten external frameworks arrive at similar conclusions). We do not yet have outcome validity (proof that the framework produces the developmental results it's designed for). Moving from design validity to outcome validity is the next critical step, and it requires partners.

→ Pilot Measurement Toolkit → Development Markers overview → Engage With This Work (collaboration opportunities)

How We Iterate

Systematic Revision

Building a comprehensive framework means managing continuous change across hundreds of interconnected documents. steamHouse maintains version control and systematic revision practices:

Terminology management. Eleven formal terminology updates have been executed across the project, each with documented rationale, find-and-replace patterns, and verification checklists. When a term changes — and they do, because precision matters — the change propagates systematically rather than haphazardly. Examples: "tribe" → "supertribe" → careful contextual usage; book title "WIRED" → "HOOKED"; "Stone Age Minds" → "Wired for Yesterday" (book title).

Document versioning. Major documents carry version numbers and change logs. The Investment Case is at v2. The Terminology Master Update is at v11. The Book Suite inventory is at v8. These aren't arbitrary numbers — each version reflects substantive revision in response to new evidence, partner feedback, or design refinement.

Cross-reference integrity. When a structural decision changes — such as the Globe Team evolving from a two-book to a three-book architecture — the implications cascade through multiple documents. We track these cascades explicitly and execute updates in documented sequences.

Why this matters to you: If you're evaluating steamHouse materials, you're looking at a living system, not a static product. The materials are better now than they were six months ago, and they'll be better six months from now. The iteration rate is itself evidence of seriousness — this isn't a one-time design that's being defended, but a framework under active refinement.