{ "$schema": "https://pointcast.xyz/for-agents", "code": "UES-205", "slug": "collective-intelligence-studio", "title": "Collective Intelligence Studio: Humans, Agents & Editorial Judgment", "status": "active", "frame": "Build human-agent workflows in which tools expand the field of possibility while people remain accountable for sources, decisions, omissions, and the finished public claim.", "outcome": "A sourced research dossier or creative work accompanied by a legible process ledger showing what the learner, tools, and sources each contributed.", "path": "/ues/collective-intelligence-studio", "jsonPath": "/ues/collective-intelligence-studio.json", "weeklyCommitment": { "selfGuidedStudioMinutes": 75, "fieldworkMinutes": { "minimum": 60, "maximum": 90 }, "reflectionMinutes": 20, "total": "About 2.5–3 hours per module" }, "access": [ "Every instruction is readable on the public course page; audio or video references require a transcript or equivalent text path.", "Every required prompt, scenario, rubric, and checklist is written on this course page; named sources are optional references, not missing course materials.", "Low-bandwidth, print-friendly syllabus with no paid platform or subscription required.", "No live attendance, camera, account, wallet, or public posting is required.", "Progress stays in this browser unless the learner chooses to download a private completion receipt.", "No AI system is required. A complete no-model route uses search, index cards, manual variations, and the same source and editorial checks." ], "materials": [ "Browser and text editor", "Notebook with separate columns for claim, source, tool or method, candidate output, human edit, uncertainty, and decision", "Optional AI system the learner already chooses to use; paper-based role cards fully replace it", "No coding experience required" ], "instructorProfile": "An editor or creative technologist with strong source verification, authorship, privacy, model-literacy, and collaborative production practices.", "soloPath": "Run researcher, tool operator, and editor as distinct passes. Keep the process ledger honest about each role; no team or paid model subscription is required.", "assignments": [ { "title": "Prompt and source notebook", "dueWeek": 2, "brief": "Keep claims, sources, prompts, model outputs, human edits, uncertainties, and discarded paths in distinct fields." }, { "title": "Human-agent process map", "dueWeek": 4, "brief": "Diagram which human pass or tool proposes, checks, decides, edits, attributes, and signs off at each stage of a one-person workflow; collaboration is optional." }, { "title": "Dossier and process ledger", "dueWeek": 6, "brief": "Package the finished work privately with sources, limitations, authorship statement, material model use, and the learner’s final sign-off. Publication is optional." } ], "weeks": [ { "week": 1, "title": "Define the Roles", "question": "Which work can a tool propose, and which decisions must a named person own?", "sourceShelf": [ "On-page role rule: researcher gathers, generator proposes, critic challenges, verifier checks evidence, editor decides, and approver signs off; one learner may perform every pass", "On-page ledger fields: time, role, input, tool or manual method, output, source, human change, uncertainty, decision, and reason", "No-model role cards: apply three constraints by hand, create three candidates, then switch roles before judging them" ], "studio": "Decompose a small project into proposing, retrieving, checking, deciding, making, editing, and approving.", "fieldTask": "Observe one existing workflow and mark where automation helps, obscures, or falsely appears authoritative.", "publicReceipt": "A first human-agent responsibility map naming the learner’s final editorial decision point; optional collaborators may be added." }, { "week": 2, "title": "Build a Trustworthy Source Trail", "question": "Can every important public claim travel back to evidence a person can inspect?", "sourceShelf": [ "On-page source test: prefer firsthand records, identify the creator and date, open the cited evidence, compare an independent source, and label conflicts", "Citation, quotation, paraphrase, and uncertainty examples", "Model hallucination and retrieval-failure cases" ], "studio": "Trace claims through search, retrieval, model summaries, and human verification without treating fluency as evidence.", "fieldTask": "Build a five-source shelf around the research question and independently verify one tool-, search-, or manually generated candidate claim from primary evidence. No AI use is required.", "publicReceipt": "A source notebook excerpt showing one confirmed, one corrected, and one unresolved claim." }, { "week": 3, "title": "Prompts Are Sketches", "question": "How can prompting widen a search without quietly deciding the answer?", "sourceShelf": [ "On-page variation exercise: make one factual, one counterfactual, and one constraint-heavy approach; without AI, write the three candidate outlines yourself before editing", "Creative constraint and editorial brief examples", "Image, text, and code provenance prompts" ], "studio": "Run divergent prompts, compare outputs, annotate assumptions, and write an editorial brief before selecting material.", "fieldTask": "Generate or manually draft three meaningfully different approaches and document why the learner’s editorial pass rejects at least one.", "publicReceipt": "A prompt triptych with output differences, editorial notes, and material model disclosure." }, { "week": 4, "title": "Design the Handoffs", "question": "Where does information get lost when a project moves between people and agents?", "sourceShelf": [ "On-page gate for every pass: define the input, expected output, evidence check, access check, stop condition, saved version, and who or which role decides", "Structured brief, schema, and handoff patterns", "Version, source, and approval ledger examples" ], "studio": "Turn an improvised one-person process into explicit inputs, outputs, checks, stop conditions, and self-attested decision gates; optional team handoffs may be mapped too.", "fieldTask": "Run one simulated self-handoff between two named passes and record ambiguity, duplication, and failure points. An optional collaborator handoff uses the same standard and counts equally.", "publicReceipt": "The revised human-agent process map and one concrete rule added after the test." }, { "week": 5, "title": "Red-Team the Work", "question": "What would make this work misleading, harmful, derivative, private, or wrong?", "sourceShelf": [ "Bias, privacy, fabrication, and sensitive-data review prompts", "Copyright, style imitation, and synthetic-media disclosure cases", "Adversarial fact-check and counterexample methods" ], "studio": "Run the safety-and-rights pass in the solo toolkit, then try to disprove every major claim, find one counterexample, identify a missing affected perspective, and mark what would make the work unsafe to release. An exchanged dossier is optional.", "fieldTask": "Resolve or explicitly label every high-risk issue in the self-generated red-team report. Optional peer findings use the same standard.", "publicReceipt": "A red-team disposition list: fixed, disclosed, removed, or unresolved." }, { "week": 6, "title": "Publish the Ledger", "question": "Can another person understand how the work was made and where judgment entered?", "sourceShelf": [ "Public process-ledger examples", "Model and dataset disclosure fields", "Use the source-trail and handoff-package fields printed in the solo toolkit" ], "studio": "Complete editorial review, source audit, authorship statement, limitation note, and correction route.", "fieldTask": "Package the final dossier with its process ledger and record the learner’s final self-attestation. If optional collaborators are named, their approval is needed only for their contributions.", "publicReceipt": "A private-by-default work package with inspectable sources, process ledger, limitations, credits, and correction note; publication is optional." } ], "archiveWorkNumber": "253", "archiveCategory": "AI + COLLECTIVE", "connectsFrom": [ "UES-106", "UES-107", "UES-108" ], "budgetUsd": 3300, "addedProductionNeed": "Optional faculty demonstrations, accessible process-ledger examples, and independent curriculum maintenance; the class does not depend on model or API credits.", "boundary": "No sensitive personal data, deceptive synthetic media, unlicensed imitation, or unsupported factual publication.", "human": "https://pointcast.xyz/ues/collective-intelligence-studio", "catalog": "https://pointcast.xyz/ues", "fundingProgram": "https://pointcast.xyz/university-of-el-segundo", "participation": { "status": "active", "deliveryMode": "self-paced", "start": "anytime", "durationModules": 6, "learnerPriceUsd": 0, "requirements": { "account": false, "application": false, "deadlines": false, "liveAttendance": false, "wallet": false } }, "progressPolicy": { "storage": "local-browser-only", "syncsToServer": false, "identityStored": false, "learnerArtifactsStored": false, "note": "Only completion choices and timestamps are saved in this browser. The learner may reset them or download a private receipt." }, "optionalSupportedCohort": { "availableWhenFunded": true, "requiredForAccessOrCompletion": false, "note": "A scheduled cohort may add human review and facilitation, but it never gates the independent path." }, "completionReceipt": { "kind": "self-attested", "privateByDefault": true, "walletRequired": false, "onChain": false, "disclaimer": "The receipt is not academic credit, accreditation, verified identity, an on-chain credential, or a financial credential." }, "transactionPolicy": "Course browsing and participation are wallet-free. Tezos support is optional and requires separate explicit Mainnet review." }