When Staying Calm Wins in Repeated Fights

In the paper's words

Repeated strategic encounters—such as public debates, corporate negotiations, and online discussions—are frequently marred by personal attacks that raise coordination costs, erode trust, and impede long‑term cooperation. Existing literature treats “ball‑first” (issue‑directed) conduct as an ethical recommendation but lacks a formal game‑theoretic justification. We model the interaction as an infinitely repeated symmetric two‑strategy game that extends the classic Hawk‑Dove framework by adding a reputational penalty d for aggression against a cooperative opponent.

The abstract, verbatim; every line below is quoted from the text.

What it claims

that ball‑first is ESS when $d > V/2$—would be falsified if empirical observations of repeated interactions with measured V, C, and d showed a persistent prevalence of person‑first aggression despite d exceeding V/2. A concrete falsification test is described in Section 8.3.

Numbers it reports

1. The only admissible equilibria are the boundaries, and the sign of $\dot{x}$ for $0\lt x\lt 1$ is positive, driving the system toward $x=1$ only because the term $0.9x$ originates from the $C$ parameter; the absence of $d$ eliminates the penalty that would otherwise protect B, illustrating the necessity of a non‑zero reputational cost for the ball‑first policy to be evolutionarily stable.

   from 7. Limiting case: vanishing penalty

2. The dynamics then admit an interior mixed equilibrium where $f_B(x^{*})=f_P(x^{*})$; for example, with $d=0.4$, $f_P(x)=x(1-0.4)+(1-x)(1-2)/2=1.1x-0.5$, and setting $0.5=1.1x^{*}-0.5$ gives $x^{*}=1/1.1\approx0.909$.

   from 4.4 Sensitivity to $d$

3. The replicator equation then simplifies to $$ \dot{x}=x\bigl(0.5-(0.9x-0.5)\bigr)=x(1-0.9x), $$ which has an interior fixed point at $x^{*}=1/0.9\approx1.11$, outside the feasible interval.

   from 7. Limiting case: vanishing penalty

4. With the concrete parameterisation $V=1$, $C=2$, $d=0.6$, the numerical threshold $d^{*}=0.5$ is exceeded, confirming that ball‑first (B) is an ESS.

   from Abstract

5. As a purely hypothetical illustration with assumed numbers, a six‑month delay of a product launch whose time‑value we assume to be $ 24 M per year of earlier availability would correspond to an assumed cost of $ 12 M; no empirical case is claimed here.

   from 1. Introduction

6. Derivation of the ESS condition $d \gt V/2$ and explicit numerical threshold $d^{*}=0.5$ for $V=1$ (Section 4).

   from 1. Introduction

How it is done

1. Using discrete replicator dynamics with a one‑step delay, we derive a closed‑form evolutionarily stable strategy (ESS) condition $d \gt V/2$, where $V$ is the resource value.

   from Abstract

2. The algorithm described for computing ESSs in symmetric normal‑form games with three or more players shows that the support‑enumeration technique can be applied to our ball‑first (B) / person‑first (P) setting when a third identical agent participates in each round [4].

   from 1. Extending the two‑strategy Hawk‑Dove model to three players

3. The support‑enumeration algorithm for multiplayer ESSs confirms that the linear penalty $d$ remains the decisive parameter when the game is extended beyond two players [4].

   from 9. Summary of cross‑disciplinary connections

4. A laboratory‑scale repeated‑game experiment could test the model: Design: Recruit participants to play a repeated Hawk‑Dove‑style game where the payoff matrix includes a configurable reputational penalty $d$ (implemented as point deductions for aggressive moves).

   from 8.3 Minimal de‑risking experiment

What would falsify it

1. The main claim—that ball‑first is ESS when $d \gt V/2$—would be falsified if empirical observations of repeated interactions with measured $V$, $C$, and $d$ showed a persistent prevalence of person‑first aggression despite $d$ exceeding $V/2$.

   from 7.2 Falsification criterion

2. A concrete falsification test is described in Section 8.3.

   from 7.2 Falsification criterion

3. Decision‑makers—policy designers, platform moderators, and negotiation facilitators—would benefit from a principled rule that predicts when clean, issue‑directed engagement (ball‑first, B) outperforms personal aggression in the long run.

   from 1. Introduction

Where the claim stops

1. Homogeneous penalty: The model assumes a single scalar reputational penalty $d$ identical for all agents; real societies may exhibit heterogeneous penalties.

   from 7.1 Limitations

2. No network structure: The well‑mixed assumption neglects spatial or network heterogeneity that can influence strategy spread (cf. [3]).

   from 7.1 Limitations

3. As a purely hypothetical illustration with assumed numbers, a six‑month delay of a product launch whose time‑value we assume to be $ 24 M per year of earlier availability would correspond to an assumed cost of $ 12 M; no empirical case is claimed here.

   from 1. Introduction

Where to read

1. Abstract176 words · 1 min
2. 1. Introduction280 words · 1 min
3. 2. Background and Related Work172 words · 1 min
4. 3. Methods291 words · 1 min
5. 4. Analysis464 words · 2 min
6. 5. Results80 words · 1 min
7. 6. Extended Derivations and Worked Examples1,169 words · 5 min
8. 7. Discussion238 words · 1 min
9. 8. Practical Implications and Engineering Pathways352 words · 2 min
10. 9. Conclusion113 words · 1 min
11. Provenance and AI Involvement225 words · 1 min