Skip to the content.

Scope and Strategic Roadmap

Computational Foundations for Strategic Coopetition: Research Program Architecture

This document provides the theoretical rationale, scope justification, and strategic roadmap for the ongoing research program on computational modeling of strategic coopetition. It serves as the authoritative reference for guiding and directing future extensions and advancements.


Document Purpose

This scope roadmap addresses fundamental questions about the research program’s design: 1. Why does coopetition-gym use a [0, endowment] action space?

  1. How is this grounded in game theory and coopetition literature?
  2. What is the relationship between cooperation and competition in the model?
  3. How will the research program evolve to address broader treatments?

The document establishes that the current implementation represents Phase 1 (Foundations) of a multi-phase research program, with Phase 2 (Extensions) planned to introduce biaxial treatments of cooperation and competition.


Part I: The Coopetition Modeling Debate

1.1 Defining Coopetition

Coopetition is a portmanteau of “cooperation” and “competition” describing the phenomenon of simultaneous cooperation and competition between economic actors. The central maxim of coopetition is:

“Cooperate to grow the pie, compete to split the pie.” , Brandenburger & Nalebuff (1996)

This implies that cooperation and competition may operate as distinct strategic dimensions rather than opposite poles of a single continuum.

1.2 The Uniaxial vs. Biaxial Debate

The coopetition literature contains a substantive debate about whether cooperation and competition should be modeled as:

Uniaxial (Continuum) Treatment

Authors Work Position
Bengtsson & Kock (2000) “Coopetition in Business Networks” Cooperation-competition as ends of a spectrum; firms position along continuum
Lado, Boyd & Hanlon (1997) “Competition, Cooperation, and the Search for Economic Rents” “Syncretic rent-seeking” behavior on single dimension
Quintana-García & Benavides-Velasco (2004) “Cooperation, Competition, and Innovative Capability” Coopetition intensity as single variable measure
Padula & Dagnino (2007) “Untangling the Rise of Coopetition” Coopetition as intermediate position between pure cooperation and pure competition

Key argument: Firms must allocate limited resources between cooperative and competitive activities; more of one implies less of the other. The choice is fundamentally about positioning on a cooperation-competition spectrum.

Biaxial (Orthogonal) Treatment

Authors Work Position
Brandenburger & Nalebuff (1996) “Co-opetition” Value creation (cooperation) vs. value capture (competition) as distinct activities
Luo (2007) “Coopetition in International Business” Two dimensions: “coopetition intensity” × “coopetition scope”
Gnyawali & Park (2011) “Co-opetition between Giants” Simultaneous high cooperation AND high competition possible and observed
Bengtsson, Kock & Johansson (2014) “A Systematic Review of Coopetition” Recognizes both unidimensional and multidimensional conceptualizations
Bouncken et al. (2015) “Coopetition: A Systematic Review” Identifies tension between “balance” (uniaxial) and “simultaneity” (biaxial) views

Key argument: Firms can simultaneously invest heavily in both cooperation (joint R&D, standard-setting) and competition (pricing, market positioning). These are orthogonal strategic choices.

1.3 Implications for Computational Modeling

Aspect Uniaxial Treatment Biaxial Treatment
Action space 1D: [0, max] or [-max, +max] 2D: [0, max]_coop × [0, max]_comp
Strategic tension Resource allocation tradeoff Independent optimization
Nash equilibrium Single dimension Multi-dimensional
Computational tractability Higher (1D optimization) Lower (2D optimization)
MARL compatibility Direct (standard continuous) Requires adaptation
Interpretability Clear single metric Richer but complex

Both treatments are legitimate modeling choices with different strengths. The debate is not resolved in the literature, and different research questions may warrant different treatments.


Part II: Foundations Series Justification (TR-1 through TR-4)

2.1 Scope of the Foundations Series

The “Computational Foundations for Strategic Coopetition” technical reports (TR-1 through TR-4) adopt the uniaxial treatment following the social dilemma tradition. This is a deliberate and defensible modeling choice.

Technical Report Focus Action Interpretation
TR-1: Interdependence & Complementarity Value creation through joint investment a_i = contribution to synergy
TR-2: Trust & Reputation Dynamics Trust evolution from observed cooperation a_i = cooperation signal
TR-3: Collective Action & Loyalty Team production and loyalty a_i = effort contribution
TR-4: Sequential Interaction & Reciprocity History-dependent cooperation a_i = reciprocal response

2.2 Theoretical Justification

The Foundations series models the cooperation dimension of coopetition, with competition captured through structural parameters:

Competition via Structural Parameters

Parameter Symbol Competitive Interpretation
Bargaining shares α_i Value capture allocation
Interdependence matrix D_ij Structural power asymmetries
Trust dynamics T_ij, R_ij Reputation-mediated future access
Endowment asymmetries e_i Resource-based competitive advantage

This approach reflects the insight that competition often manifests through structural position rather than explicit competitive actions. An agent with high α captures more value regardless of action choice; an agent with asymmetric D_ij (high dependency on them, low dependency on others) holds structural power.

Grounding in Game Theory Literature

The [0, endowment] action space is standard in established game-theoretic models:

Game Class Action Space “Zero” Meaning Literature
Public Goods [0, E] Contribute nothing Ledyard (1995), Isaac et al. (1988)
Trust Game [0, E] Send nothing Berg, Dickhaut & McCabe (1995)
Continuous PD [0, 1] Full defection Killingback & Doebeli (2002)
Social Dilemmas [0, max] Selfish action Leibo et al. (2017), MARL literature
Common Pool Resources [0, harvest_max] Restrain extraction Ostrom et al. (1992)

The Foundations series extends this tradition to model cooperation with structural competition, not “cooperation vs. no action.”

2.3 What the Current Model Captures

Phenomenon How Modeled Where
Value creation Synergy function g(a) TR-1, all environments
Value capture Bargaining shares α All environments
Power asymmetry Interdependence matrix D PartnerHoldUp-v0, PlatformEcosystem-v0
Trust dynamics T, R evolution TR-2, TrustDilemma-v0
Reputation effects Reputation damage ceiling TR-2, RecoveryRace-v0
Free-riding Nash equilibrium analysis TR-3, TeamProduction-v0
Loyalty Welfare internalization TR-3, LoyaltyTeam-v0
Coalition stability Entry/exit dynamics TR-3, CoalitionFormation-v0
Reciprocity History-dependent response TR-4, ReciprocalDilemma-v0, GiftExchange-v0, IndirectReciprocity-v0, GraduatedSanction-v0, AppleAppStore-v0

2.4 What the Current Model Does NOT Capture

The uniaxial treatment does not model active competition as a strategic choice:

Competitive Mechanism Example Status in Foundations
Price competition Bertrand undercutting Not modeled
Quantity competition Cournot flooding Not modeled
Sabotage Actively harming rival Not modeled
Market capture Aggressive positioning Not modeled
Rent-seeking Contest for fixed prize Not modeled

These mechanisms require agents to take actions that impose direct costs on rivals,something the [0, endowment] action space cannot express. This is the domain of the planned Extensions series.

2.5 Terminology Clarification

To avoid confusion, the following terminology mapping is adopted:

Common Term Foundations Meaning More Precise Term
“Defection” Non-contribution to joint value Retention, Non-investment
“Full cooperation” Maximum contribution to synergy Full investment
“Cooperation level” Contribution to value creation Investment level
“Competitive dynamics” Structural position effects Structural competition

The term “defection” is borrowed from Prisoner’s Dilemma literature and should be understood as “selfish retention” rather than “active harm.”


Part III: Extensions Series Roadmap (TR-5 onwards)

3.1 Planned Transition to Biaxial Treatment

The Extensions series will introduce biaxial action spaces where cooperation and competition are independent strategic dimensions.

FOUNDATIONS (TR-1 to TR-4)          EXTENSIONS (TR-5 onwards)
         │                                    │
    Uniaxial                              Biaxial
         │                                    │
    a ∈ [0, E]                    a = (a_coop, a_comp)
    Cooperation                   a_coop ∈ [0, E_coop]
    dimension                     a_comp ∈ [0, E_comp]
    only                          Both dimensions

3.2 Planned Technical Reports

Report Title Focus Key Innovation
TR-5 Value Capture & Market Competition Active competition dimension Biaxial action space
TR-6 Sabotage & Defensive Strategies Negative externality actions Harm modeling
TR-7 Multi-Market Contact & Spillovers Cross-market dynamics Portfolio competition
TR-8 Coalition Competition & Rivalry Inter-coalition competition Nested game structure

3.3 TR-5: Value Capture & Market Competition (Planned)

Focus: Introduce explicit competition dimension alongside cooperation.

Biaxial Action Space Design

# Proposed action space for TR-5 environments
action_space = spaces.Dict({
    "cooperation": spaces.Box(low=0.0, high=100.0, shape=(1,), dtype=np.float32),
    "competition": spaces.Box(low=0.0, high=100.0, shape=(1,), dtype=np.float32)
})

# Alternative: Single Box with two dimensions
action_space = spaces.Box(
    low=np.array([0.0, 0.0]),
    high=np.array([100.0, 100.0]),
    shape=(2,),
    dtype=np.float32
)

Reward Structure

# Biaxial reward function (conceptual)
def biaxial_reward(actions, params):
    # Cooperation creates value
    total_value = synergy_function(actions["cooperation"])

    # Competition allocates value
    shares = competition_function(actions["competition"])

    # Individual payoff
    payoff_i = shares[i] * total_value

    # Competition has direct costs
    competition_cost = cost_function(actions["competition"][i])

    return payoff_i - competition_cost

Planned Environments

Environment Description Key Feature
ValueCapture-v0 Basic biaxial environment Independent coop/comp dimensions
PriceCompetition-v0 Bertrand-style pricing Price undercutting dynamics
MarketShare-v0 Quantity competition Market flooding effects
R&D-Race-v0 Innovation competition Technology capture

3.4 TR-6: Sabotage & Defensive Strategies (Planned)

Focus: Actions that actively harm rivals (negative externalities).

Tripolar Action Space

# Tripolar action space: cooperate, compete, or sabotage
action_space = spaces.Box(
    low=np.array([0.0, 0.0, 0.0]),
    high=np.array([100.0, 100.0, 100.0]),
    shape=(3,),
    dtype=np.float32
)
# Dimensions: [cooperation, competition, sabotage]

Sabotage Mechanics

Action Type Effect on Self Effect on Rival
Cooperation (a_coop) + Synergy share + Synergy share
Competition (a_comp) + Larger share - Smaller share
Sabotage (a_sab) - Direct cost - Output reduction

Planned Environments

Environment Description Key Feature
Sabotage-v0 Basic sabotage mechanics Negative externality actions
DefensiveInvestment-v0 Protection against sabotage Defense allocation
IndustrialEspionage-v0 Information sabotage Knowledge appropriation

3.5 TR-7 and TR-8 (Conceptual)

Report Focus Novel Contribution
TR-7 Multi-market contact Cross-market retaliation and forbearance
TR-8 Coalition competition Intra-coalition cooperation, inter-coalition competition

Part IV: Environment Architecture Evolution

4.1 Current Environment Categories (v1.x)

Category Environments Treatment
Dyadic TrustDilemma-v0, PartnerHoldUp-v0 Uniaxial
Ecosystem PlatformEcosystem-v0, DynamicPartnerSelection-v0 Uniaxial
Benchmark RecoveryRace-v0, SynergySearch-v0 Uniaxial
Case Study SLCD-v0, RenaultNissan-v0, ApacheProject-v0, AppleAppStore-v0 Uniaxial
Extended CooperativeNegotiation-v0, ReputationMarket-v0 Uniaxial
TR-3 TeamProduction-v0, LoyaltyTeam-v0, CoalitionFormation-v0, ApacheProject-v0, PublicGoods-v0 Uniaxial
TR-4 ReciprocalDilemma-v0, GiftExchange-v0, IndirectReciprocity-v0, GraduatedSanction-v0, AppleAppStore-v0 Uniaxial

4.2 Planned Environment Categories (v2.x)

Category Environments Treatment
Biaxial Basic ValueCapture-v0, BiaxialDilemma-v0 Biaxial
Market Competition PriceCompetition-v0, MarketShare-v0 Biaxial
Technology R&D-Race-v0, PatentCompetition-v0 Biaxial
Sabotage Sabotage-v0, DefensiveInvestment-v0 Tripolar
Multi-Market CrossMarket-v0, PortfolioCompetition-v0 Biaxial
Coalition Rivalry CoalitionRivalry-v0, AllianceWar-v0 Nested biaxial

4.3 Backward Compatibility

All v1.x environments will be preserved in v2.x releases:

# v2.x usage - both paradigms available
import coopetition_gym

# Uniaxial environments (Foundations)
env_uniaxial = coopetition_gym.make("TrustDilemma-v0")
# action_space: Box([0, 100], shape=(2,))

# Biaxial environments (Extensions)
env_biaxial = coopetition_gym.make("ValueCapture-v0")
# action_space: Box([[0,0], [100,100]], shape=(2, 2))

References

Coopetition Literature (Uniaxial Tradition)

Coopetition Literature (Biaxial Tradition)

Game Theory Foundations

Multi-Agent Reinforcement Learning

Technical Reports (This Research Program)

  1. Pant, V., & Yu, E. (2025). Computational Foundations for Strategic Coopetition: Formalizing Interdependence and Complementarity. arXiv:2510.18802.

  2. Pant, V., & Yu, E. (2025). Computational Foundations for Strategic Coopetition: Formalizing Trust and Reputation Dynamics. arXiv:2510.24909.

  3. Pant, V., & Yu, E. (2025). Computational Foundations for Strategic Coopetition: Formalizing Collective Action and Loyalty. arXiv:2601.16237.

  4. Pant, V., & Yu, E. (2026). Computational Foundations for Strategic Coopetition: Formalizing Sequential Interaction and Reciprocity. arXiv:2604.01240. Link


Document History

Version Date Changes
1.0 2026-02-09 Initial scope roadmap documenting Foundations justification and Extensions roadmap