Ignacio Adrián Lerer Independent Researcher · Buenos Aires, Argentina ORCID: 0009-0007-6378-9749 Email: [email protected] Zenodo communities: law-as-extended-phenotype · small-concept-models Substack: adrianlerer.substack.com

Ignacio Adrián Lerer is a lawyer (Universidad de Buenos Aires) and Executive MBA (IAE Business School, Universidad Austral) based in Buenos Aires, Argentina. He is an independent researcher working at the intersection of evolutionary biology, game theory, institutional economics, and law.

His research program applies Dawkins's (1982) Extended Phenotype Theory and Dennett's intentionality hierarchy to legal institutions, producing a suite of quantitative diagnostic instruments validated against empirical datasets from four jurisdictions. He maintains 136 Zenodo DOI records and a historical SSRN archive of 62 records, with all SSRN archive entries cross-referenced to their Zenodo records. SSRN currently exposes only a subset of that historical archive, so this repository and Zenodo are the authoritative preservation sources. He has one article published and one manuscript under review at the Journal of Computational Law and Legal Technology (JCLLT).

He is not affiliated with a university. All work is independent research.

The papers/ directory is the canonical source set for the research-program NotebookLM notebook. After adding or replacing a PDF, sync missing sources with:

LERER_NOTEBOOKLM_ID=<notebook-id> python3 tools/sync_notebooklm_papers.py --dry-run
LERER_NOTEBOOKLM_ID=<notebook-id> python3 tools/sync_notebooklm_papers.py

The script requires the nlm CLI from adrianlerer/notebooklm-mcp-cli and uploads only filenames that are not already present as NotebookLM source titles.

Legal norms function as cultural replicators (memes) subject to Darwinian selection. Their fitness is defined as:

Fitness = P(transmission) × P(compliance) × P(enforcement)

Institutions are the extended phenotypes these replicators build for their own reproduction: specialized courts, university curricula, bar associations, enforcement agencies. These structures persist long after the original norms they served have been formally repealed.

The program integrates five theoretical traditions that have not previously been combined:

1. Dawkins's Extended Phenotype Theory (1982) — the causal reach of replicators extends beyond the organism into the environment
2. Dennett's Intentionality Hierarchy (1987) — agents operate at different levels of intentional complexity (L0–L3+)
3. Evolutionary Game Theory (Maynard Smith 1982, multilevel selection) — ESS analysis, replicator dynamics, Price equation
4. Hayek's Spontaneous Order (1973) — institutional patterns emerge without central design
5. Scott's Infrapolitics (1985, 1990) — hidden transcripts and everyday resistance as selection mechanisms

Hard core (non-negotiable):

• Legal norms are cultural replicators subject to variation, selection, and retention
• Institutions are extended phenotypes of dominant memeplexes
• Multilevel selection operates in legal systems (intra-group and inter-group)
• Agents operate at asymmetric intentionality levels (L1 corporations vs. L3 humans)
• Legal professionals learn norms by observing authoritative reactions, not empirical outcomes (HBU)

Protective belt (empirically adjustable):

• Specific CLI component weights
• IHR numerical thresholds
• Dennett-Nash Gap empirical estimates

Positive heuristic (expansion directions):

• New jurisdictions (UK, US, Japan, India, South Africa)
• New domains (tax law, environmental law, AI regulation, IP)
• Computational implementation (CriptoIus, JurisRank, IusMorfos)

Measures institutional rigidity ex ante (before reform attempt):

CLI = 0.30×P + 0.25×D + 0.20×O + 0.25×E

• P = Precedent density (citation network entrenchment)
• D = Doctrinal elaboration (academic consensus volume)
• O = Professional organization (bar certification, specialized courts)
• E = Enforcement infrastructure (administrative capacity)

Validation: R²=0.74, AUC=0.97, 60 labor reform cases, 4 jurisdictions (Argentina, Brazil, Chile, Spain). ICC=0.9335 inter-rater reliability (4 LLMs, n=30 US constitutional provisions).

Key empirical results:

Counterintuitive finding: crisis reinforces lock-in in high-CLI systems (interaction term β=−2.83, p<0.05). Political economy variables lose significance when CLI is controlled.

Prospective validation: Milei 2024 reform program — CLI predicted 12.4% success probability; observed 0% (p=0.003). CGT injunction against Law 27.802 (March 30, 2026) confirmed as observation #24: blocking interval of 24 days, Bayesian posterior updated from P(blocking)=0.92 to 0.94 (95% CI: 0.78–0.99). See 10.5281/zenodo.19340088. The computational validation framework is published in JCLLT: 10.47852/bonviewJCLLT62027951.

Thresholds:

• CLI < 0.30: flexible (ordinary legislation works)
• CLI 0.30–0.50: reformable with difficulty (supermajority + crisis)
• CLI 0.50–0.70: severe lock-in (requires constitutional amendment)
• CLI > 0.70: requires regime change

Measures irreversible institutional deformation ex post (after reform failure). Scale 0–1.

Distinction from CLI: CLI is structural (ex ante), IHR is historical (ex post, includes infrastructure of formally repealed norms). Argentina Art. 14bis: IHR ≈ 0.87.

Zombie Law Equilibrium: when CLI > 0.85 AND Implementation Gap > 0.60 simultaneously — formal reform blocked, informal bypass individually rational, stable equilibrium.

Design criterion for building evolvable institutions. Four components (Kirschner & Gerhart 2005):

• V (Variation): capacity to generate normative variants = f(1−CLI)
• S (Selection): efficacy of outcome-based selection mechanism
• C (Conservation): capacity to retain successful innovations
• M (Modularity): capacity to reform subsystems without fragmenting the whole

Predicted values: Chile IEI > 0.70, Argentina IEI < 0.30.

Intentionality levels (Dennett 1987):

• L1: behavior predicted by attributing first-order goals — corporations, algorithms, RL agents
• L3: behavior predicted by attributing recursive beliefs about beliefs — adult humans in strategic contexts

Asymmetric Intentionality Theory (AIT): Social coordination mechanisms (shame, reputation, reciprocity) require common knowledge that presupposes L3+. L1 agents are not merely unresponsive — they are structurally immune.

Dennett-Nash Gap:

Δᵢ = SAᵢ + ICSᵢ − MPᵢ + SIPᵢ

• SA = Strategic Advantage (exploiting false symmetry assumption)
• ICS = Intentionality Cost Savings (cognitive overhead avoided)
• MP = Mismatch Penalty (losses from misattributing mental states)
• SIP = Social Immunity Premium (structural immunity to social sanctions)

Empirical estimates: Tech ToS 15–48% | Argentine executive decrees 40–62% | Sovereign debt ~40pp

P_A(H|E) = [P(E|H)^w(A) × P(H)] / P(E)    where w(A) > 1

Legal professionals (and AI systems trained with RLHF) learn norms by observing authoritative reactions (R_a: authority convergence; R_i: institutional elaboration; R_s: coordinated sanctions; R_n: narrative entrenchment) — not empirical outcomes. This generates reform-resistant priors.

Argentina/Brazil comparison: ratio of perceptual to active inference 41:1 (Argentina) vs. 9:1 (Brazil).

Dawkins-Hayek synthesis: order emerges without central design (Hayek), but the entities generating it optimize for their own fitness, not systemic efficiency (Dawkins). Three activation conditions: visible inequality, organized intermediaries, weakened enforcement.

Six failure modes in human-institutional interaction:

1. Strategic exploitation (SA dominant)
2. Cognitive cost asymmetry (ICS dominant)
3. Communication breakdown (MP dominant)
4. Social mechanism failure (SIP dominant)
5. Aumann impossibility (when min(Lᵢ, Lⱼ) < 3, posteriors cannot become common knowledge)
6. Dynamic Classification Failure (intentionality level changes during interaction — e.g., AI systems transitioning from L1 to L2 during deployment)

Modified PageRank over judicial citation networks with temporal decay. Validated: ρ=0.73, p<0.001, 1,247 Argentine Supreme Court cases. Detects doctrinal crossover points 3–5 years before political visibility.

• CriptoIus: blockchain-based legal precedent system as evolutionary constraint for multi-agent AI governance
• IusMorfos: ex ante prediction of reform success probability from CLI components
• RootFinder: inverse genealogical tracing through citation networks
• JurisRank: PageRank-based doctrinal fitness scoring
• GitHub: github.com/adrianlerer

All papers are freely downloadable. Communities: law-as-extended-phenotype · small-concept-models.

The SSRN account was suspended in January 2026 for administrative reasons (category mismatch for interdisciplinary papers), unrelated to content. SSRN availability is now partial and unstable: as of 2026-07-04, the author page exposed 43 public records out of the historical 62. The supporting snapshot and extracted audit are preserved in library/ssrn-audits/. The 19 records missing from the 2026-07-04 author-page snapshot are: 5768423, 5881663, 5881702, 5882342, 5886142, 5891884, 5908462, 5908662, 5955735, 5960677, 5971914, 5974555, 5991115, 5991195, 5991355, 5995234, 5995435, 6000675, and 6001434. This table preserves the complete historical SSRN archive. All PDFs are hosted in the papers/ folder of this repository, and Zenodo links identify the durable DOI record when available. For preservation and citation, use the GitHub-hosted PDF and Zenodo DOI as the authoritative sources.

Lerer, I.A. (year). [Title]. [SSRN Working Paper ID / Zenodo DOI].
Available: https://github.com/adrianlerer/lerer-research-program