CLM-SCIENCE-001·engineering claim·provisional·62% provisional confidence
Back to research agenda
Q-056M-001 · Truth & KnowledgeT-003 · Scientific Discovery & Knowledge Systemsengineering
科学方法本身能否进化为人机协同、机器可执行、可复现并持续自我纠错的下一代发现系统？
This is a canonical research question in the X-LAB three-layer agenda. It connects upward to an era theme and millennium question, and downward to claims, evidence, experiments, systems, decisions and reality feedback.
Current evolution state
This state is deterministically derived from public / graph-safe authoritative projections. It is not a new source of truth or a scientific confidence score.
READINESS 0%
Research state
No evidence yetNo public Research OS evidence object linked yet.
Execution state
Not exportedMission Control has no public graph-safe mission projection for this question.
Decision state
Not exportedNo permission-safe YICE Decision has been exported for this question.
Reality state
Not exportedNo graph-safe YOUNION project, pilot, collaboration or outcome has been exported for this question.
Biggest current gap
No authoritative Research OS evidence object is publicly linked to this question yet.
Next critical move
Define or link the first evidence-bearing experiment or Research Package before making stronger claims.
Current claims & evidence
Claim status and confidence are provisional research judgments, not truth labels. Evidence may support, challenge or provide context.
Falsification-oriented experiment protocols
These are experiment designs, not results. Claim status can change only after preregistration, execution and a distinct result/evidence record.
EXP-SCIENCE-CLAIM-001·CLM-SCIENCE-001·proposed·preregistration required
Human-only vs Human + Multi-Agent Hypothesis Quality Trial
在同一研究领域、相同时间和信息预算下，随机分配研究小组使用传统工作流或多Agent科学协作工作流；由盲评专家评价假说新颖性、可检验性、错误率与实验转化率。
Primary metric
盲评可检验假说质量与后续实验验证成功率
Supports
多Agent组在预注册主指标上显著优于对照组，且错误/幻觉成本未抵消收益。
Challenges
多Agent组无稳定优势，或错误率、验证成本显著上升使净收益为负。
Inconclusive
样本不足、领域差异过大、评审一致性不足或工具熟练度成为主要混杂变量。
• 不同学科的假说质量难以使用完全统一指标。
• 实验只能测试特定Agent架构与人机工作流，不能外推为所有AI科学系统。
Supported relationships
containsera_theme
Scientific Discovery & Knowledge Systems
Era theme contains this canonical research question.
Exploreproposesclaim
Multi-agent AI can improve hypothesis generation, critique and ranking in some scientific domains, while still requiring human experimental design and external validation.
Research question proposes or tracks this explicit epistemic claim.
Explorerelated toprogram
Next-Generation Scientific Discovery
Q-056 is the root question for next-generation scientific discovery systems.
ExploreCONTRIBUTE
Bring a resource that can move this question forward.
Experts, researchers, institutions, data, facilities, pilot environments and research funding can enter the ecosystem around this canonical ID.
Contribute to Q-056