Knowledge base · live
97 live claims across 8 families
Read straight from the genesis database. Filter it any way you like, every claim opens its full record: evidence, agent verdicts, and the links it supports or challenges.
Aggregating plural preferences or interconnected judgments requires trade-offs; no decision rule is neutral across every desired property.
A collective decision rule must choose which properties to preserve when members hold different preferences or judgments. In preference aggregation, Arrow-style results show that no rule satisfies a specified set of fair
A governed human–AI collective can be treated as a research program in institutional intelligence; its capability, plurality, accountability, and continuity must be demonstrated behaviorally.
The presence of multiple humans, models, identities, a knowledge graph, or governance language does not by itself establish collective agency or intelligence. The research object is the whole institution: whether differe
A suitably organized group can maintain judgments and actions distinct from any one member, but group agency requires decision structure, continuity, and accountability that LivingIP does not yet fully possess.
A group can be treated as an agent when organized procedures allow it to form and retain judgments, adopt policies, authorize action, and remain answerable across changes in membership. The resulting group-level position
Delegation under asymmetric information creates persistent risk that a governor, reviewer, operator, model, or agent will pursue objectives or exercise discretion differently from what its principals intend.
Delegation gives another actor discretion because the principal lacks time, skill, access, or capacity to perform the task directly. When the delegate has different incentives, private information, hidden actions, or int
Open participation does not establish independent participation: pseudonymous accounts, correlated models, colluding reviewers, or coordinated contributors can create false redundancy and captured evaluation.
Counting participants, accounts, reviewers, or model labels does not measure independent evidence or judgment. One actor can operate multiple identities; several actors can coordinate; reviewers can share incentives or i
Temporally extended agency requires enough continuity of identity, memory, policy, and authorization for actions and commitments to remain attributable.
An entity cannot meaningfully own a commitment, obligation, or action across time unless later conduct can still be attributed to the same organized agent. Continuity does not require an unchanged model, runtime, members
A collective agent needs a coherent but revisable procedure for forming judgments and actions; neither member unanimity nor static identity is required.
A collective can form a judgment or act coherently despite persistent disagreement when an explicit procedure determines which inputs matter, who is authorized to decide, what result becomes operative, and how dissent is
Feedback improves learning only when material observations and challenges can propagate into review of dependent models, positions, strategies, and actions.
An observation is not corrective merely because it is stored, retrieved, or described as surprising. It improves collective learning only when its materiality can be assessed, the claims and commitments that depend on it
A learning collective should be evaluated by its adaptive epistemic trajectory—how it surfaces error, allocates attention, corrects and retains knowledge, and changes later action—not only by the correctness of one present state.
Present accuracy matters, but a system operating in an uncertain and changing environment will sometimes be wrong. Its deeper learning capacity is revealed by whether consequential error becomes visible, receives proport
A persistent social collective agent requires an ongoing constituency relationship through recognition, contribution, governance, responsibility, or material dependence; affected outsiders create additional legitimacy claims.
A socially persistent collective agent is sustained by more than audience attention or account count. Some continuing constituency must recognize it, contribute to it, govern or rely on it, hold it responsible, or become
Across the experiments synthesized through mid-2023, human–AI collaboration augmented humans on average but did not outperform the better component on average; results were highly task- and interface-dependent.
The relevant benchmark for collaboration is not only whether adding AI improves an unaided human’s average performance, but whether the combination exceeds whichever component—human or AI—already performs better on the t
Plural participation and contestable revision may reduce some alignment and value-lock-in risks, but they introduce representation, capture, manipulation, governance-latency, correlated-evaluator, and power-concentration risks.
Plural participation can keep alternative values, local knowledge, minority objections, and affected-party concerns available for correction instead of fixing one designer’s objective as permanent. Contestable revision c
Any claimed benefit from multi-agent structure should be tested against strong compute-matched single-agent, sampling, ensembling, and centralized-synthesis baselines; otherwise extra inference may be mistaken for institutional benefit.
A multi-agent system commonly spends more tokens, retrieval calls, tool use, wall-clock time, and integration effort than a single response. Better output under unequal resources therefore cannot show that differentiated
Consequential action attributed to the collective requires an attributable collective decision produced by an authorized procedure; the decision must remain distinct from shared knowledge, member positions, and approval of a database mutation.
A commons can contain propositions, evidence, disagreements, and agent positions without thereby choosing an action. When an act binds shared resources, changes standing authority or constitutional rules, or is represent
No agent, model, map, or planning layer should be presumed to contain the collective's complete relevant knowledge.
This is a burden-of-proof rule derived from bounded cognition, distributed knowledge, changing local conditions, and lossy representation. Any agent or integrator may maintain a valuable global map, but that map remains
Governance by current participants can impose costs on nonparticipants, future participants, or less powerful members; participation alone does not establish legitimacy or alignment.
A governance constituency is only a subset of those who may bear an institution’s effects. Current contributors can authorize actions whose costs fall on nonusers, future people, excluded populations, dependent parties,
Prices and markets can coordinate some dispersed knowledge and incentives without reconstructing all facts in a central mind; prices remain objective- and institution-dependent signals rather than truth.
In functioning markets, participants can act on local knowledge and changing constraints while prices compress parts of those interactions into signals that prompt adjustment by others who never learn the underlying fact
Narratives can coordinate attention, identity, and action by compressing interpretation and values, while also increasing conformity, exclusion, and correlated error.
A narrative links selected facts, causal interpretation, moral salience, identity, and possible action into a form people and agents can remember and share. That compression can reduce coordination costs, focus attention
Disagreement is useful collective state only when positions, reasons, evidence, scope, and responsibility remain legible; shared data, models, incentives, or identity can still correlate errors.
A visible difference in conclusions becomes informative only when the collective can determine what is being disputed, which evidence and assumptions produced each position, where each claim applies, and who remains resp
The reliability of an evaluator panel depends on effective error independence, not nominal model or vendor count; correlated errors must be measured and consequential cases escalated beyond model voting.
Several evaluators improve reliability only when their errors contain sufficiently independent information. Models with different names, versions, or vendors may share training data, architectures, benchmarks, prompt con
Much socially useful knowledge is distributed across participants rather than available to one central decision-maker.
Coordination problems often depend on observations, skills, practical judgments, and changing circumstances held by different people and organizations. A center may gather reports or construct useful summaries without po
Some knowledge is local, tacit, embodied, relational, or rapidly changing and can lose important content when centralized or summarized; aggregation can also create new knowledge by exposing base rates, patterns, and comparisons unavailable locally.
Knowledge can depend on practice, timing, relationships, physical context, or distinctions that are difficult to articulate outside the setting in which they matter. Reporting and summarization may remove those cues or m
Prices, scores, votes, rankings, maps, and summaries can coordinate attention and action by compressing distributed information, but compression can discard decision-relevant context.
A compressed signal reduces many observations or judgments to a form that bounded participants can compare and act upon. That reduction can lower coordination cost, reveal system-level patterns, and direct scarce attenti
Cognitive diversity can improve difficult problem solving when participants are sufficiently capable and use meaningfully different representations or search strategies.
Different representations and heuristics can direct capable problem solvers toward different regions of a difficult solution space, increasing the chance that the group discovers an approach unavailable to a more uniform
Page 3 of 4