Integrating behavioral experimental findings into dynamical models to inform social change interventions

Manuel S. Mariani; Radu Tanase; Ren\'e Algesheimer

arxiv: 2405.13224 · v2 · submitted 2024-05-21 · ⚛️ physics.soc-ph · cs.SI· econ.EM

Integrating behavioral experimental findings into dynamical models to inform social change interventions

Radu Tanase , Ren\'e Algesheimer , Manuel S. Mariani This is my paper

Pith reviewed 2026-05-24 01:05 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cs.SIecon.EM

keywords complex contagiondiscrete choice modelingseeding policiesbehavior adoptionsocial networksadoption thresholdsdynamical models

0 comments

The pith

Integrating individual adoption thresholds from experiments into network models shows standard seeding policies for behavioral change are suboptimal.

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

The paper combines discrete choice experiments with models of complex contagion to estimate personal thresholds for adopting new behaviors or products. It validates the estimates in two experiments and then inserts them into network simulations. These simulations indicate that seeding strategies based only on network position can be improved once individual decision drivers are measured and included. A reader would care because large-scale adoption drives responses to challenges such as sustainable consumption or health practices. The work supplies a concrete experimental route to bring individual-level data into dynamical network models that previously treated thresholds as uniform or absent.

Core claim

By integrating discrete choice modeling into complex contagion theory the authors estimate individual-level thresholds to adoption. They validate the predictive power of this approach in two choice experiments. When the estimated thresholds are placed into computational simulations of social networks, state-of-the-art seeding policies prove suboptimal if they neglect individual-level behavioral drivers, and the experimental method supplies a correction.

What carries the argument

Estimation of individual adoption thresholds via discrete choice experiments, then inserted into dynamical simulations of complex contagion on networks.

If this is right

Seeding policies informed by measured individual thresholds produce higher large-scale adoption than policies that ignore them.
Controlled experiments can supply thresholds that improve the accuracy of network-based adoption forecasts.
Interventions for social change can be adjusted by correcting for the individual behavioral drivers that pure network models omit.

Where Pith is reading between the lines

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

The same experimental-plus-simulation pipeline could be applied to design public campaigns for vaccination or energy conservation.
Models might next examine how thresholds shift when network neighbors exert repeated influence rather than one-shot exposure.
Repeating the experiments across different behaviors could reveal whether the size of the correction varies by domain.

Load-bearing premise

Thresholds measured in controlled choice experiments apply to real social-network adoption without large distortion from context or network structure.

What would settle it

A field test in which seeding guided by the experimentally measured thresholds produces lower or equal final adoption than seeding guided by standard network-centrality rules alone.

Figures

Figures reproduced from arXiv: 2405.13224 by Manuel S. Mariani, Radu Tanase, Ren\'e Algesheimer.

**Figure 1.** Figure 1: Estimating individual–level thresholds. (A) Threshold-based diffusion models assume that individual–level adoption choices are determined by the decision-makers’ threshold. On the other hand, individual-level perspectives focus on estimating individuals’ utilities of adopting from choice data. Utility-based approaches to behavioral change reconcile the two perspectives by reinterpreting the individual-leve… view at source ↗

**Figure 2.** Figure 2: Relative performance of seeding policies. (A) Nodes selected by different policies in an illustrative network structure: The highest-degree node (in blue) has the largest number of connections, but the highest-neighborhood susceptibility node (in dark red) has the largest number of connections to low-threshold nodes (in red). (B, C) Relative performance of seeding policies under a preference–based cost str… view at source ↗

**Figure 1.** Figure 1: Estimating individual–level thresholds. (A) [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗

**Figure 2.** Figure 2: Relative performance of seeding policies. (A) [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗

read the original abstract

Addressing global challenges often involves stimulating the large-scale adoption of new products or behaviors. Research traditions that focus on individual decision making suggest that achieving this objective requires identifying the drivers of individual discrete adoption choices. On the other hand, computational approaches rooted in complexity science focus on maximizing the propagation of a given product or behavior throughout social networks of interconnected adopters. Here, by integrating discrete choice modeling into the complex contagion theory, we propose a method to estimate individual-level thresholds to adoption. We validate the predictive power of this approach in two choice experiments. By integrating the estimated thresholds into computational simulations, we show that state-of-the-art seeding policies for initiating large-scale behavioral change might be suboptimal if they neglect individual-level behavioral drivers, which can be corrected through the proposed experimental method.

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 integration of discrete choice experiments into complex contagion models is a reasonable step but the transfer of thresholds to network dynamics is untested.

read the letter

The main thing to know is that this paper estimates adoption thresholds from discrete choice experiments and plugs them into complex contagion simulations, then argues that standard seeding policies may be suboptimal because they ignore those individual drivers. The experiments validate the choice model on its own terms, and the simulations illustrate the difference in outcomes when thresholds vary by individual factors rather than uniform assumptions. That combination is the actual new piece: using experimental data to set the parameters instead of deriving them from network observations or assuming fixed values. The experiments themselves look like a straightforward check on predictive accuracy for individual decisions, which is useful as far as it goes. The simulations are consistent with the inputs they receive. The soft spot is the mapping step. Thresholds recovered from isolated binary choices are inserted as neighbor-fraction rules in the network model, but nothing tests whether those numbers survive when the decision context includes actual social exposure, repeated interactions, or network structure. The validation stays inside further choice experiments, so the simulations only demonstrate internal consistency under the assumption that the mapping holds. That leaves the policy claim about suboptimal seeding resting on an unverified transfer. If the thresholds shift once embedded in real networks, the simulation results do not carry over. This is aimed at computational social scientists who want to ground network models in behavioral data. It has enough structure and a clear method to merit referee time, though the external validity of the thresholds will need direct attention. I would send it for peer review.

Referee Report

2 major / 1 minor

Summary. The paper integrates discrete choice modeling into complex contagion theory to estimate individual-level adoption thresholds from behavioral experiments. It validates this approach in two choice experiments and incorporates the thresholds into network simulations to demonstrate that state-of-the-art seeding policies for behavioral change may be suboptimal when individual behavioral drivers are neglected.

Significance. If the mapping from experimental thresholds to network contagion dynamics is valid, the work bridges individual decision-making research with complexity science, offering a method to improve intervention design for large-scale behavioral adoption. The experimental validation of threshold estimation is a positive feature.

major comments (2)

[Validation and simulation integration sections] The validation of the threshold estimation approach is performed only within additional choice experiments; the manuscript provides no direct test (e.g., via network experiments or out-of-sample network simulations) that these thresholds correspond to the neighbor-fraction adoption rule used in the complex contagion model. This mapping is load-bearing for the claim that seeding policies are suboptimal.
[Methods and simulation sections] The central construction assumes that thresholds recovered from isolated discrete-choice tasks equal the behavioral thresholds agents would apply when the sole varying input is the realized fraction of adopting neighbors, without distortion from network structure, repeated exposure, or contextual effects; this assumption is not empirically tested beyond the choice-experiment setting.

minor comments (1)

[Methods] Clarify the exact functional form used to translate choice-experiment utilities into neighbor-fraction thresholds for the simulation update rule.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We appreciate the referee's detailed review and the opportunity to clarify aspects of our work. The comments highlight important considerations for the empirical grounding of our integrated modeling approach. We respond to each major comment below and indicate where revisions will be made to the manuscript.

read point-by-point responses

Referee: [Validation and simulation integration sections] The validation of the threshold estimation approach is performed only within additional choice experiments; the manuscript provides no direct test (e.g., via network experiments or out-of-sample network simulations) that these thresholds correspond to the neighbor-fraction adoption rule used in the complex contagion model. This mapping is load-bearing for the claim that seeding policies are suboptimal.

Authors: We acknowledge that our validation is conducted within the discrete choice experiment framework rather than through direct network experiments. The experiments were structured to present participants with varying fractions of adopters to directly elicit the threshold parameter used in the complex contagion model. This provides a controlled measurement of the individual-level threshold. We agree that out-of-sample validation in network settings would be valuable. Accordingly, we will revise the manuscript to include a dedicated paragraph in the Discussion section discussing the scope of the current validation and outlining directions for future empirical tests in networked environments. This revision clarifies the claims without altering the presented results. revision: partial
Referee: [Methods and simulation sections] The central construction assumes that thresholds recovered from isolated discrete-choice tasks equal the behavioral thresholds agents would apply when the sole varying input is the realized fraction of adopting neighbors, without distortion from network structure, repeated exposure, or contextual effects; this assumption is not empirically tested beyond the choice-experiment setting.

Authors: This assumption is central, and we recognize it is tested only in the isolated choice setting. The design of the experiments aims to isolate the effect of the adoption fraction by holding other factors constant, providing a baseline for the threshold. Potential distortions from network structure or repeated exposure are important but would require a different experimental paradigm. We will update the Methods section to more explicitly state this assumption and its rationale, and add robustness checks in the simulation results to examine sensitivity to variations in thresholds that might arise from such effects. We believe these changes address the concern by increasing transparency. revision: partial

Circularity Check

0 steps flagged

No significant circularity; thresholds sourced from independent experiments

full rationale

The paper estimates adoption thresholds via discrete-choice experiments and inserts the resulting values as fixed inputs into complex-contagion simulations. No equations are presented that derive those thresholds from the dynamical model itself, nor are any load-bearing steps shown to reduce to self-citations, fitted parameters renamed as predictions, or ansatzes smuggled via prior author work. Validation occurs inside further choice experiments, and the simulations test policy consequences of the externally measured thresholds rather than recovering them by construction. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only; no free parameters, axioms, or invented entities are described.

pith-pipeline@v0.9.0 · 5667 in / 920 out tokens · 20806 ms · 2026-05-24T01:05:08.650345+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[1]

Calibration of partworth utilities.Each agent is characterized by ann par-dimensional vector whose elements constitute the partworth utilities. Thusn par = 24 in the PS study andn par = 13 in the AA study (one parameter for each level of each attribute, one parameter for the social signal, and two parameters for the none option). To calibrate the choice s...

work page
[2]

Generation of choice tasks.We generateZ= 15 choice tasks per agent, each with a choice set consisting of two product alternatives and a none option. Each product alternative in the choice set is described as a unique combination of the product attributes from the focal conjoint study, and one out of three levels of the social signal: 0.1, 0.5 or 0.9. For ...

work page
[3]

activated network

Choice simulation.We simulate the conjoint survey by simulating the choices made by the agents. To simulate which alternative was selected in each choice set, we compute the utilities corresponding to each alternative, and the utility of the none option. For each agent, we compute the utility of a given alternative by summing up the attribute utilities fo...

work page

[1] [1]

Calibration of partworth utilities.Each agent is characterized by ann par-dimensional vector whose elements constitute the partworth utilities. Thusn par = 24 in the PS study andn par = 13 in the AA study (one parameter for each level of each attribute, one parameter for the social signal, and two parameters for the none option). To calibrate the choice s...

work page

[2] [2]

Generation of choice tasks.We generateZ= 15 choice tasks per agent, each with a choice set consisting of two product alternatives and a none option. Each product alternative in the choice set is described as a unique combination of the product attributes from the focal conjoint study, and one out of three levels of the social signal: 0.1, 0.5 or 0.9. For ...

work page

[3] [3]

activated network

Choice simulation.We simulate the conjoint survey by simulating the choices made by the agents. To simulate which alternative was selected in each choice set, we compute the utilities corresponding to each alternative, and the utility of the none option. For each agent, we compute the utility of a given alternative by summing up the attribute utilities fo...

work page