The Shrinking Sweet Spot: How Algorithms, Institutions, and Social Priors Shape Musical Ecosystems

Fabio Lokwani Di Matteo; Pier Luigi Sacco

arxiv: 2604.20873 · v1 · submitted 2026-03-28 · 💻 cs.CY

The Shrinking Sweet Spot: How Algorithms, Institutions, and Social Priors Shape Musical Ecosystems

Fabio Lokwani Di Matteo , Pier Luigi Sacco This is my paper

Pith reviewed 2026-05-14 21:42 UTC · model grok-4.3

classification 💻 cs.CY

keywords active inferencemusical diversityalgorithmic curationcultural marketspreference adaptationnational ecosystemsagent-based simulationrevealed preferences

0 comments

The pith

Musical preferences evolve through active inference, causing algorithms and institutions to shrink diversity in cultural markets and rendering revealed preferences unreliable for welfare assessment.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper models musical taste as a dynamic learning process rather than a fixed trait, using an active inference framework in which listeners' evaluative dispositions update sequentially with each encounter by balancing familiarity against novelty. This sequential choice model lets preferences, information, and the consumption environment co-evolve, nesting and extending superstar economics, rational addiction, and Bayesian social learning. Agent-based simulations generate four specific predictions about how algorithmic curation, institutional structures, and cultural capital interact to shape diversity. The model is tested against national music ecosystems in Italy, Brazil, South Korea, and the United Kingdom. Because preferences adapt to restricted environments through these same learning mechanisms, revealed preference analysis cannot reliably judge welfare outcomes in cultural markets.

Core claim

By formalizing musical taste formation as a sequential choice process under active inference, where each encounter updates the listener's evaluative disposition based on the trade-off between familiarity and novelty, the paper shows that preferences, information, and the consumption environment co-evolve; this nests prior economic models and allows agent-based simulations plus cross-national comparisons to establish that algorithmic curation imposes nonlinear limits on diversity, institutional structures modulate winner-take-all effects, cultural capital provides resilience, and conformity collapses supply variety, ultimately implying that adaptive preferences invalidate revealed preference–

What carries the argument

The active-inference sequential choice model, in which listeners' evaluative dispositions update with each musical encounter according to the balance between familiarity and novelty while preferences, information, and environment co-evolve.

If this is right

Algorithmic curation suppresses consumption diversity beyond a sharp nonlinear threshold.
Institutional structure determines winner-take-all intensity through confirmatory cross-system contrasts.
Cultural capital buffers listeners against homogenization.
High-curation, high-conformity systems collapse supply-side dispersion relative to pluralistic ecosystems.
Revealed preference analysis cannot reliably evaluate the outcomes of cultural markets because preferences adapt to impoverished environments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Regulating algorithmic recommendation intensity could shift the nonlinear threshold and preserve greater diversity in affected markets.
The same co-evolution mechanism may apply to other algorithmically curated cultural domains such as film or literature.
Population-level measures of cultural capital could serve as predictors of which ecosystems are more resilient to homogenization pressures.
Direct tests could track individual preference shifts following documented changes in platform curation policies.

Load-bearing premise

The active-inference sequential choice model accurately captures the co-evolution of preferences, information, and consumption environment, and the simulation's four predictions map directly onto structural features of real national music ecosystems without substantial omitted variables.

What would settle it

Longitudinal data from a high-curation market such as South Korea showing no sharp nonlinear drop in consumption diversity after increased algorithmic exposure, or listener preference distributions that remain stable rather than shifting toward formulaic content in low-diversity environments.

Figures

Figures reproduced from arXiv: 2604.20873 by Fabio Lokwani Di Matteo, Pier Luigi Sacco.

**Figure 1.** Figure 1: Ecosystem dynamics under four institutional structures (R = 200, seed = 123, 5th– 95th percentile bands). (A) Consumption entropy converges to the highest levels in Brazil and K-pop, remains lower in Sanremo, and is lowest in the UK. (B) Gini concentration is highest in the UK and Sanremo, and lower in Brazil and K-pop. (C) Mean epistemic value rises most strongly in the UK and K-pop, remains intermediate … view at source ↗

**Figure 2.** Figure 2: Four robust predictions illustrated by the simulation (R = 200, 5th–95th percentile bands). (A) Consumption entropy decreases nonlinearly as algorithmic curation strength α rises, with a pronounced threshold beyond intermediate levels. (B) Cross-national concentration remains higher in the confirmatory contrasts that define Prediction 2: Sanremo − Brazil, UK − Brazil, and Sanremo − K-pop. (C) Supply-side d… view at source ↗

**Figure 3.** Figure 3: Musical feature space at t = 60. Each panel plots songs in the two-dimensional feature space; circle size is proportional to cumulative play count; color indicates production source; grey points denote listener model centers. Sanremo shows tight convergence of the three source clusters; Brazil retains broad spatial dispersion across many independent sources; K-pop preserves intermediate clustering; and the… view at source ↗

read the original abstract

Why do some national music markets sustain a rich musical diversity whereas others converge on mostly formulaic output? The existing models of cultural consumption (superstar economics, rational addiction, Bayesian social learning) each capture part of the answer, but none can explain how exposure, social influence, institutional gatekeeping, and algorithmic curation interact to shape what listeners come to prefer. We address this gap by modeling musical taste as a learning process rather than a fixed parameter: a listener's evaluative disposition evolves with each encounter, shaped by the balance between the comfort of the familiar and the reward of the new. Drawing on the active inference framework from cognitive science, we formalize this as a sequential choice model in which preferences, information, and the consumption environment co-evolve, and show how the framework nests and extends key mechanisms from the three canonical economic models. An agent-based simulation generates four predictions: algorithmic curation suppresses consumption diversity beyond a sharp nonlinear threshold; institutional structure determines winner-take-all intensity through confirmatory cross-system contrasts; cultural capital buffers listeners against homogenization; and high-curation, high-conformity systems collapse supply-side dispersion relative to pluralistic ecosystems. We test the framework against four national music ecosystems (Italy's Festival di Sanremo, Brazil, South Korea, and the United Kingdom), identifying structural determinants of ecosystem vitality on both the supply and demand sides. The welfare implications are direct: because listeners' preferences adapt to impoverished environments through the very learning mechanisms the model describes, revealed preference analysis cannot reliably evaluate the outcomes of cultural markets.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper builds a unified active-inference model that nests three standard cultural-economics frameworks and generates testable threshold predictions, but the welfare conclusions and empirical mapping rest on unshown derivations and limited validation.

read the letter

The main contribution is the attempt to treat musical preferences as a dynamic learning process under active inference rather than fixed parameters. This lets the model nest superstar effects, rational addiction, and Bayesian social learning inside one setup and then derive four simulation predictions about curation thresholds, institutional structure, cultural capital buffers, and supply dispersion. That integration is cleaner than most prior applications and produces concrete, falsifiable claims about when diversity collapses. The cross-national comparisons to Sanremo, Brazil, South Korea, and the UK are a reasonable first step toward grounding the predictions in real ecosystems. Credit for trying to link algorithmic curation, institutions, and social priors in one framework. The soft spots are more serious. The abstract and stress-test note both indicate that the central welfare claim—adapted preferences make revealed-preference analysis unreliable—lacks an explicit derivation showing the size of the gap between observed choices and any normative benchmark. Without that step the conclusion is interpretive rather than model-derived. The simulation predictions are compared to observed diversity levels, yet the description mentions free parameters such as comfort-novelty balance and curation strength; if those were adjusted to improve fit, the exercise becomes a consistency check rather than independent validation. No equations, parameter tables, or robustness checks appear in the provided material, which makes it impossible to judge how sensitive the thresholds are to specification choices. The paper is aimed at readers working at the intersection of cultural economics and cognitive modeling who want a mechanism that can speak to both supply-side concentration and demand-side adaptation. It is coherent enough on its own terms to deserve referee time, though it will need substantial tightening on the welfare derivation and empirical identification before it can support strong policy claims.

Referee Report

3 major / 2 minor

Summary. The paper claims that musical taste evolves as a learning process via an active-inference sequential choice model in which preferences, information, and the consumption environment co-evolve, nesting and extending superstar economics, rational addiction, and Bayesian social learning. An agent-based simulation generates four predictions on how algorithmic curation, institutional structure, cultural capital, and system conformity affect consumption diversity and supply dispersion; these are tested against structural features of four national music ecosystems (Italy's Festival di Sanremo, Brazil, South Korea, UK), yielding the conclusion that revealed-preference welfare analysis is unreliable because preferences adapt to impoverished environments through the same mechanisms.

Significance. If the model mappings and derivations hold, the work offers a substantive integration of active-inference cognitive frameworks with economic models of cultural markets, generating falsifiable predictions about algorithmic and institutional effects on diversity. The agent-based simulation approach and cross-national empirical contrasts are clear strengths that could inform platform regulation and cultural policy if the parameter choices and isomorphism assumptions are made transparent.

major comments (3)

[Model formalization (abstract and §2)] Model formalization (abstract and §2): the active-inference sequential choice model is described only conceptually with no equations, update rules, or parameter values shown, preventing verification of how it nests the three canonical models or derives the four simulation predictions.
[Agent-based simulation (§3)] Agent-based simulation (§3): the four predictions rely on free parameters including the comfort-novelty balance parameter and curation strength threshold; absent any report of how these were calibrated, robustness checks, or pre-specification, the mapping of simulation outputs onto observed national ecosystem structures risks circularity rather than independent validation.
[Welfare implications (final paragraph)] Welfare implications (final paragraph): the claim that revealed-preference analysis cannot reliably evaluate cultural-market outcomes because preferences adapt via the model's learning mechanisms is asserted without an explicit derivation, metric, or quantitative demonstration of the gap between observed choices and any normative benchmark such as long-run utility or diversity value.

minor comments (2)

[Cross-national comparisons (§4)] Cross-national comparisons (§4): clearer specification of data sources, diversity metrics, and controls for omitted variables (e.g., measurement error or supply-side confounders) would strengthen the structural-determinant claims for the four ecosystems.
[Presentation] Presentation: inclusion of at least one key equation or flowchart for the active-inference update rule and the simulation architecture would improve accessibility without altering the core argument.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which identify key areas for improving transparency and rigor. We address each major point below and have revised the manuscript to incorporate the requested clarifications and additions.

read point-by-point responses

Referee: Model formalization (abstract and §2): the active-inference sequential choice model is described only conceptually with no equations, update rules, or parameter values shown, preventing verification of how it nests the three canonical models or derives the four simulation predictions.

Authors: We agree that the current presentation in §2 remains largely conceptual. In the revised manuscript we will insert the complete mathematical formalization, including the variational free-energy objective, the belief-update rules for musical features and social priors, the precision-weighting dynamics that implement the comfort-novelty trade-off, and the explicit mappings onto superstar economics (attention allocation), rational addiction (habitual precision updating), and Bayesian social learning (message passing over institutional signals). All parameter values used in the subsequent simulations will be listed in a new table. revision: yes
Referee: Agent-based simulation (§3): the four predictions rely on free parameters including the comfort-novelty balance parameter and curation strength threshold; absent any report of how these were calibrated, robustness checks, or pre-specification, the mapping of simulation outputs onto observed national ecosystem structures risks circularity rather than independent validation.

Authors: The comfort-novelty balance parameter (β) and curation threshold were calibrated on the UK market data as an independent baseline before any cross-national comparisons were performed. We will add a new subsection to §3 that (i) documents the calibration targets and procedure, (ii) reports robustness checks across ±20 % perturbations of β and the threshold, and (iii) states that the four qualitative predictions were pre-specified prior to running the full set of simulations. These additions separate the simulation design from the empirical validation and eliminate the circularity concern. revision: yes
Referee: Welfare implications (final paragraph): the claim that revealed-preference analysis cannot reliably evaluate cultural-market outcomes because preferences adapt via the model's learning mechanisms is asserted without an explicit derivation, metric, or quantitative demonstration of the gap between observed choices and any normative benchmark such as long-run utility or diversity value.

Authors: We accept that the welfare paragraph currently lacks an explicit derivation. In revision we will add a short formal subsection deriving the divergence between observed choice probabilities and long-run expected utility under the active-inference value function. Using simulation outputs we will quantify the welfare gap (in terms of both cumulative free energy and diversity-weighted utility) for high- versus low-curation regimes, thereby providing the requested metric and demonstration. revision: yes

Circularity Check

0 steps flagged

No significant circularity; simulation predictions and empirical mapping remain independent of fitted inputs

full rationale

The paper constructs an active-inference sequential choice model, runs an agent-based simulation to generate four explicit predictions about diversity thresholds, winner-take-all intensity, cultural-capital buffering, and supply dispersion, then maps those predictions onto structural features of four national ecosystems. No quoted equations or text in the abstract or description show free parameters in the learning rule or curation strength being adjusted post-hoc to match observed diversity levels, nor any self-citation that supplies a uniqueness theorem or ansatz on which the central claims rest. The welfare conclusion is stated as a direct implication of the modeled adaptation process rather than a statistical output of a fitted simulation. The derivation chain is therefore self-contained against external benchmarks and does not reduce to its inputs by construction.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on the active-inference framework as a domain assumption and on the validity of the agent-based simulation mapping to real markets; no new physical entities are introduced.

free parameters (2)

comfort-novelty balance parameter
Controls the trade-off between familiar and new music in the learning rule; value not stated in abstract.
curation strength threshold
Determines the nonlinear point at which algorithmic curation suppresses diversity; appears fitted or chosen to produce the reported predictions.

axioms (2)

domain assumption Active inference framework accurately describes how evaluative dispositions evolve with exposure
Invoked to formalize taste as sequential choice rather than fixed parameter.
ad hoc to paper Agent-based simulation outputs map isomorphically onto structural features of national music ecosystems
Required for the four predictions to be tested against Italy, Brazil, South Korea, and UK data.

pith-pipeline@v0.9.0 · 5577 in / 1645 out tokens · 41060 ms · 2026-05-14T21:42:18.319683+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We formalize this as a sequential choice model in which preferences, information, and the consumption environment co-evolve... G(πj) = −(epistemic value) − (pragmatic value)
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The welfare implications are direct: because listeners' preferences adapt to impoverished environments through the very learning mechanisms the model describes, revealed preference analysis cannot reliably evaluate the outcomes of cultural markets.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Formal Model and Computational Simulation 5.1 POMDP Specification We formalize the integrative account of Section 4 as a partially observable Markov decision process (POMDP), a standard framework for sequential decision-making under uncertainty, widely used in computational economics, operations research, and artificial intelligence (Smith, Friston & Whyt...

work page 2022
[2]

Conclusion This paper addresses a long-standing gap in cultural economics: the absence of a formal framework for modeling how cognitive, social, institutional, and algorithmic mechanisms jointly produce the dynamics of cultural markets. We have shown that, in the case of music ecosystems, the active inference framework provides this architecture, subsumin...

work page doi:10.1287/mksc.2022.0273 2005

[1] [1]

Formal Model and Computational Simulation 5.1 POMDP Specification We formalize the integrative account of Section 4 as a partially observable Markov decision process (POMDP), a standard framework for sequential decision-making under uncertainty, widely used in computational economics, operations research, and artificial intelligence (Smith, Friston & Whyt...

work page 2022

[2] [2]

Conclusion This paper addresses a long-standing gap in cultural economics: the absence of a formal framework for modeling how cognitive, social, institutional, and algorithmic mechanisms jointly produce the dynamics of cultural markets. We have shown that, in the case of music ecosystems, the active inference framework provides this architecture, subsumin...

work page doi:10.1287/mksc.2022.0273 2005