A six-point epistemic architecture turns ambiguous fossil traces into testable hypotheses by forcing explicit assumptions, alternative enumeration, bounded inference and robustness tests; applied to AI economics, the same checklist would tighten causal attribution, standardize metrics and raise policy-relevant evidence standards.

Main Finding

The paper argues that palaeontology — and ichnology in particular — benefits from an explicit epistemic architecture: a set of logical criteria that make inference transparent, testable and repeatable. Applying six formal criteria to trace-based interpretation (e.g., burrows, tracks) transforms competing narratives into testable hypotheses and strengthens claims about tracemaker identity, behaviour and environment.

Key Points

• Motivation

  • Current palaeontological reasoning often relies on implicit judgements linking material evidence to biological interpretation; this reduces reproducibility and logical rigor.
  • Ichnology is a useful testbed because of high ambiguity: a single organism can make diverse traces, and similar traces can arise from different organisms or abiotic processes.

• Six proposed criteria (conceptual scaffold)

• Clarity — make concepts, assumptions and inferential steps explicit and precisely defined.
• Necessity and sufficiency — specify which observations are required (necessary) and which combinations would be decisive (sufficient) for a given interpretation.
• Completeness — ensure the hypothesis space covers all plausible alternative explanations and relevant variables.
• Boundedness and normalization — define the spatial, temporal and observational bounds of inference and normalize measurements so comparisons are meaningful.
• Convergence — expect independent lines of evidence (morphology, sedimentology, geochemistry, modern analogues) to converge on the same inference; use convergence as weight of evidence.

• Consistency and robustness — test whether inferences hold under different methods, scales, assumptions and perturbations (sensitivity analyses).

• Practical questions addressed

  • Are existing criteria adequate to infer a bilaterian origin for burrows?
  • How does scale affect measurement of bioturbation and heterogeneity, particularly from the tracemaker’s perspective?
  • Can universal scaling laws link trace morphology, organismal biology and environmental conditions?
  • How can ichnotaxonomy, ethology (behavioural inference) and tracemaker affinity be made more logically rigorous?

• Outcome

  • The paper shows how applying these criteria to representative ichnological cases reveals weaknesses in common reasoning practices and suggests concrete methodological refinements.

Data & Methods

• Nature of the work
  • Conceptual and methodological paper: developing an epistemic framework rather than reporting new field data.
• Methods used to operationalize the framework
  • Formalization of inferential criteria (as above) and articulation of their logical roles in hypothesis testing.
  • Re-examination of representative ichnological case studies to illustrate application of the criteria. Examples include burrow attribution, measures of bioturbation and scaling analyses.
  • Comparative reasoning across evidence types: morphology, stratigraphy, sedimentology, taphonomy, modern analogues and potential abiogenic processes.
  • Use of robustness checks and alternative-hypothesis enumeration to expose where common inferences fail completeness, sufficiency or boundedness.
  • Discussion of how to develop universal scaling relationships (conceptual path rather than empirical derivation).

Implications for AI Economics

The paper’s central methodological prescriptions map directly to challenges in AI economics where attribution, measurement, and inference are crucial. Key implications and concrete transfers:

• Adopt an explicit epistemic architecture for causal and attributional claims about AI impacts

  • Make assumptions, inferential steps and definitions explicit (Clarity).
  • Articulate necessary vs sufficient conditions for claims like “AI caused job displacement” or “algorithmic trading caused volatility spikes.”

• Broaden and formalize hypothesis spaces (Completeness)

  • Enumerate alternative mechanisms (automation, offshoring, demand shocks, measurement artifacts) and design analyses that can discriminate among them.
  • Prevent single-explanation narratives by requiring explicit coverage of plausible alternatives.

• Define bounds and normalize metrics (Boundedness & Normalization)

  • Specify temporal and spatial bounds for inference (firm-, sector-, region-, economy-level).
  • Normalize metrics across datasets and platforms (user counts, compute-hours, model versions) to enable comparable inference.
  • Pay attention to scale: micro-level behaviors of models may not aggregate linearly to macroeconomic outcomes.

• Use convergence of independent evidence streams

  • Combine time-series econometrics, natural experiments, model-based counterfactuals, qualitative firm-level data, and synthetic controls; require convergence before strong causal claims.
  • Cross-validate findings across different datasets and identification strategies.

• Prioritize consistency and robustness testing

  • Systematically run sensitivity analyses to modeling choices, measurement error, aggregation levels and counterfactual construction.
  • Report failures of robustness as informative constraints, not just caveats.

• Develop scaling laws and standardized taxonomies

  • Analogous to trace morphology → organismal inference, build standardized taxonomies of AI system behaviors (e.g., worker-displacing automation modes, augmentation modes, market-impact strategies).
  • Seek empirical scaling relationships linking micro-level AI metrics (model size, latency, usage intensity) to macro outcomes (productivity, employment shares, price dynamics) and explicitly state their domains of validity.

• Improve reproducibility and regulatory evidence standards

  • Use the six criteria as an evaluation checklist for policy-relevant claims about AI impacts (e.g., mandatory reporting standards, evidence grading).
  • Encourage sharing of raw usage logs, model metadata and pre-analysis plans to facilitate repeatable inference.

Recommended practices for AI economists (practical checklist derived from the paper) - Make claims explicit: state necessary and sufficient indicators for each claim. - Enumerate alternative hypotheses and design tests that can falsify them. - Define the spatial/temporal/observational scope of conclusions and normalize measurements for comparability. - Combine multiple independent data sources and methods; weight inference by convergence. - Run systematic robustness and sensitivity analyses; document which inferences survive perturbations. - Work toward standardized taxonomies and empirically validated scaling relationships that link AI-system traces to economic outcomes.

Overall, the epistemic architecture proposed for ichnology provides a transferable blueprint for strengthening logical rigor, testability and policy relevance in AI economics.