Ecogame: Feedback-Driven Environmental Simulation

• Ecogame is a term covering environmentally themed games and simulations that use feedback mechanisms to couple actions with environmental outcomes.
• It spans diverse applications including environmental education, eco-evolutionary modeling, and simulation platforms for policy and behavioral change.
• Interactive feedback loops in ecogame systems modify game dynamics, offering practical insights into sustainability and resource management.

In the literature represented here, “Ecogame” denotes several related but distinct objects: environmentally themed serious games for teaching and behavior change, deliberative simulations for climate and energy transitions, eco-evolutionary games in which strategy and environment coevolve, ecosystem simulators implemented in game-engine or discrete-event settings, and, in a historically specific sense, a 1970 cybernetic art project titled Ecogame (Chiotaki et al., 2020, Konstantakopoulos et al., 2014, Bairagya et al., 2022, Strannegård et al., 2023, Mason, 9 Aug 2025, Simpson et al., 18 Aug 2025). This suggests that the term functions less as a single standardized label than as a family of feedback-centered practices in which player or population action modifies an environmental state and that altered state reorganizes subsequent choice.

1. Scope of the term

One major usage treats an ecogame as a serious game for environmental education or sustainability awareness. In this sense, the game is designed to make environmental consequences legible through scoring, rewards, comparison, narrative consequence, or repeated classroom interaction. The environmental educational card game described as a Top Trumps–style card game for second-grade learners, the office-lighting social game for energy efficiency, the AI-driven EcoEcho sustainability-awareness game, the astronomy decarbonization game “My Earth”, and the Swiss transition simulator Ensured Energy all fall into this broad category, although they differ sharply in audience, mechanics, and evidentiary goals (Chiotaki et al., 2020, Konstantakopoulos et al., 2014, Zhang et al., 2024, Malbet et al., 2024, Simpson et al., 18 Aug 2025).

A second usage is mathematical. Here an ecogame is an eco-evolutionary game or feedback-evolving game: a model in which strategy frequencies and an environmental variable are dynamically coupled. In these papers, the environment may be a renewable resource, a local cooperator multiplier, a global nonlinear factor, a common environmental stock shared by two populations, or a spatially distributed resource field. Replicator dynamics, population growth, PDEs, Bayesian incomplete information, and bargaining geometry all appear as formal variants of this meaning (Wang et al., 2020, Bairagya et al., 2022, Jiang et al., 2022, Paarporn et al., 2024, Yao et al., 10 Feb 2025, Patra et al., 3 Apr 2025).

A third usage is computational and simulation-oriented. Here the ecogame is a platform or visual simulator rather than a finite rule set. Ecotwin is presented as an open-source Unity-based ecosystem simulator with reflex, happiness, and policy networks, while a related AI-based ecosystem simulator constructs local terrains from geographic data and populates them with DRL-controlled hares and foxes. PlantProtectionSim similarly frames plant–herbivore ecology as a configurable visual discrete-event optimization game (Strannegård et al., 2021, Strannegård et al., 2023, Dietrich et al., 19 Sep 2025).

A fourth usage is historical and proper-nominal. In Catherine Mason’s account, Ecogame was an innovative British art project of 1970 that combined cybernetics, simulation, networking, and participatory decision-making to model an economy-ecology system and to show that individual behavior affects the total system (Mason, 9 Aug 2025).

2. Environmental education and behavior change

In educational practice, ecogame most often denotes a serious game that links environmental content to familiar mechanics and measurable learning outcomes. A clear example is the environmental educational card game described as a Top Trumps clone game for second-grade students around 6 years old. Its curriculum combined wildlife preservation, animals at risk, conservation status, dietary habits, and measurable attributes such as size, speed, weight, and life expectancy. The study compared three groups of 20 students over four teaching hours: a game-based experimental group, a Prezi group, and a conventional-teaching group. The reported learning gains were 8.3 for the game condition, 2.8 for the Prezi condition, and 3.3 for conventional teaching; the corresponding percentage improvements were 27.53%, 9.51%, and 10.70%. The game-experience results were also strongly positive: 19 out of 20 students (95%) reported interest and fun, 80% felt no stress, and 60% said their perception of their own ability increased (Chiotaki et al., 2020).

A different educational-behavioral design embeds the ecogame in ordinary consumption rather than in a classroom lesson. In the office-lighting social game, occupants in a real building voted for desired lighting levels in $[0,100]$, and the implemented setting was the average of all votes. Lower votes earned more points, and points were used in a lottery for three Amazon gift cards. The model formalized a comfort-versus-reward tradeoff through

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$

$$\phi_i(x_i,x_{-i})=\ln\left( \rho\frac{x_b-x_i}{nx_b-\sum_{j=1}^n x_j} \right), \qquad f_i(x_i, x_{-i})=\psi_i(x_i, x_{-i})+\theta_i\phi_i(x_i, x_{-i}).$$

Over 101 days, the office consumed 2,185 kWh for lighting and the intervention saved approximately 601 kWh, or about 27.5%. The learned Nash model also produced the best one-day-ahead prediction error among the reported baselines, with mean squared error 12.46 (Konstantakopoulos et al., 2014).

The AI-driven game EcoEcho shifts the emphasis from direct content instruction to action-consequence salience. It casts the player as KI, a scientist from 2056 who initially tries to obstruct clean energy, and uses multimodal generative-AI NPCs to prompt anti-sustainability actions whose ecological consequences then become visible. In a mixed-methods study with 23 participants, intended ecological behavior increased significantly, whereas environmental attitudes changed only slightly. The reported pre/post means were 3.40 to 4.09 for GEB, with Wilcoxon signed-rank: $Z=-3.251, p=0.001$, and 3.15 to 3.30 for NEP, with $t(22)=-1.49, p=0.15$ (Zhang et al., 2024).

3. Deliberative and institutional transition games

A more recent strand uses the ecogame as a collective planning instrument for institutional transition rather than as a direct teaching aid. In the astronomy-focused serious game “My Earth”, participants imagine themselves as members of a research laboratory who must negotiate ways to cut greenhouse-gas emissions while preserving the lab’s ability to do science. The game asks teams to build pathways that reduce emissions by 50%, focusing on space instrumentation, laboratory work, and data observation and simulation. At the Marseille workshops, seven teams and 45 participants took part, and the resulting reductions ranged from 37% to 59%. Strategies repeatedly identified as effective included replacing flights with trains for missions under 2000 km, promoting virtual meetings, extending the lifespan of scientific equipment, relying on archived data rather than organizing new observations, and pooling long-distance travel for multiple purposes (Malbet et al., 2024).

Ensured Energy applies a similar logic to national energy policy. It is an online simulation/management game embedded in a Swiss population survey, and it asks players to manage the Swiss energy system from 2022 to 2050 over 10 turns while ensuring sufficient energy provision for both summer and winter. Players can build and upgrade generation assets, import energy, attempt one policy per turn, and run political campaigns to improve policy acceptance. The game includes nuclear, hydroelectric, solar, wind, biomass, biogas, and waste, but new nuclear reactors cannot be built. It also reports explicit build and upgrade abstractions, such as Solar: build = 25%, upgrade = 5%, total upgrades = 7 and Wind: build = 100%, upgrade = 50%, total upgrades = 7. End-of-game metrics include nuclear fuel consumption, fossil fuel consumption, electricity imports, emissions (CO$_2$ equivalent), investment costs, land use, and seasonality (summer/total) (Simpson et al., 18 Aug 2025).

The survey infrastructure around Ensured Energy shows a further shift in the meaning of ecogame: from persuasion to preference elicitation. The sample drew 6,000 randomly selected Swiss residents aged 18 to 75 across all 26 cantons; 2025 respondents were assigned to treatment or control, and 1758 completed the second part of the survey. Respondents were considered “engaged” if they played at least 7 out of 10 rounds. This suggests an ecogame can function simultaneously as an information treatment, a policy-reasoning environment, and a device for observing how players resolve trade-offs under institutional and physical constraints (Simpson et al., 18 Aug 2025).

4. Eco-evolutionary and environmental-game theory

In formal game theory, ecogame usually denotes a coupled strategy-environment system. A foundational review formulates the canonical two-variable structure as

$$\epsilon \dot{x}=x(1-x)\left[ r_1(x,A(n))-r_2(x,A(n))\right],\qquad \dot{n}=n(1-n)f(x),$$

with an environment-dependent payoff matrix

$A(n)= (1-n) A_0 +n A_1.$

The paper emphasizes that cooperation may restore the environment, environmental restoration may make defection profitable, and the resulting feedback can generate an oscillating tragedy of the commons (Wang et al., 2020).

Subsequent work makes this template more ecological and more technical. For a self-renewing common resource exploited by low and high harvesters, one model couples strategy frequency $x$, total population $N$, and normalized resource $$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$0 through

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$1

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$2

Its focal regime is

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$3

for which collapse, persistence, stable limit cycles, and bistability are all possible, and finite carrying capacity can either avert or induce tragedy-of-the-commons outcomes relative to infinite-population predictions (Bairagya et al., 2022).

Other papers extend the framework by changing information, scale, or geometry. A Bayesian eco-evolutionary game introduces a noisy perception channel for the environmental state and shows that sufficiently noisy information can prevent resource extinction; a two-population model with a shared resource proves global collapse when

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$4

and a spatial PDE model with environment-driven motion shows that biased motion toward higher-quality environments can generate spatial patterns while decreasing average payoff and environmental quality across the domain (Patra et al., 3 Apr 2025, Paarporn et al., 2024, Yao et al., 10 Feb 2025).

Still further variants incorporate a global time-dependent fluctuation and a local strategy-dependent feedback, yielding an interior irregular loop in the phase plane, or use nonlinear ecological public goods games with selection-gradient-driven feedback to create multiple stable and unstable equilibrium manifolds that can be steered by switching control laws (Jiang et al., 2022, Wang et al., 2019). A different but related environmental-game usage appears in the coopetitive model of a global green economy, where a country and the rest of the world compete in biological-food production while jointly selecting an environmental policy vector

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$5

with payoffs

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$6

so that green investment and tax/incentive policy translate the Cournot payoff space (Carfì et al., 2012). The opinion-augmented model

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$7

adds a third layer in which personal opinions about the environment feed back into strategy dynamics and are themselves updated by imitation and confidence weights (Lorits et al., 2023).

5. Simulation platforms and computational ecogames

Another major meaning of ecogame is a simulation environment in which ecosystems are instantiated visually and computationally. Ecotwin is an open-source ecosystem simulator built on Unity with Unity ML-Agents. It represents space as $$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$8, and animal cognition is divided into a reflex network, a happiness network, and a policy network. The RL reward is defined directly as

$$\psi_i(x_i,x_{-i})=-\left( \bar{x}-x_i \right)^2,\qquad \bar{x}=\frac{1}{n}\sum_{i=1}^n x_i,$$9

The platform reports three studies: a wolves–deer–grass system with Lotka–Volterra-style population dynamics, a marine system in which diel vertical migration emerges, and a lethal-danger ecosystem in which agents combining RL with reflexes outperform pure RL agents (Strannegård et al., 2021).

The related AI-based ecosystem simulator for local environmental impact assessment adds geographically grounded terrain generation. It builds Unity terrains from altitude data and land cover type, encodes each animal’s local neighborhood as a $$\phi_i(x_i,x_{-i})=\ln\left( \rho\frac{x_b-x_i}{nx_b-\sum_{j=1}^n x_j} \right), \qquad f_i(x_i, x_{-i})=\psi_i(x_i, x_{-i})+\theta_i\phi_i(x_i, x_{-i}).$$0 pixel image, and controls European hare and red fox agents with PPO-trained policies. The utility and reward are

$$\phi_i(x_i,x_{-i})=\ln\left( \rho\frac{x_b-x_i}{nx_b-\sum_{j=1}^n x_j} \right), \qquad f_i(x_i, x_{-i})=\psi_i(x_i, x_{-i})+\theta_i\phi_i(x_i, x_{-i}).$$1

The simulator is explicitly intended for modeling biodiversity effects of land cover change, direct exploitation of natural resources, pollution, invasive species, and climate change, including roads, hunting, and rising sea levels (Strannegård et al., 2023).

PlantProtectionSim pushes the simulation meaning of ecogame toward configurable optimization. It is a visual discrete-event simulator for plant–herbivore interaction on grids up to

$$\phi_i(x_i,x_{-i})=\ln\left( \rho\frac{x_b-x_i}{nx_b-\sum_{j=1}^n x_j} \right), \qquad f_i(x_i, x_{-i})=\psi_i(x_i, x_{-i})+\theta_i\phi_i(x_i, x_{-i}).$$2

Plants have energy, growth, reproduction, signaling, and toxin parameters; herbivore clusters move, feed, reproduce, and can be repelled or killed. The paper states that plant energy and number of individuals in predator clusters serve as fitness measures, and it interprets the modeled ecology as a computational optimization problem inspired by nature. The reported design target is to identify settings with low energy need and long life that can cope with different patterns of attack (Dietrich et al., 19 Sep 2025).

6. Historical genealogy and documentary ambiguity

In its historically specific sense, Ecogame was a 1970 British cybernetic art system developed over about ten months by approximately 25 members of the Computer Arts Society, with George Mallen as the lead figure and principal programmer. It modeled the distribution of “wealth” through a social and industrial system “using the analogue of a reservoir of water and a plumbing system which includes a number of taps, drains and recirculating pumps.” Participants faced four-way decision points that balanced private profit against commonwealth cost, including the oil-pollution scenario in which “choice 1 yielded maximum personal profit, but imposed the maximum social cost on the commonwealth.” The installation operated over a live network linked by telephone lines to a remote time-sharing computer, used nine Tektronix graphics terminals, a large-screen interactive graphics system, joysticks, 720 35mm glass-mounted slides, and a physical water tank. It was exhibited at Computer ’70 in London, where it was played by around 5,000 people over five days, and at Davos in 1971, where it was played by delegates from 31 countries (Mason, 9 Aug 2025).

The historical paper also makes an important terminological clarification: although it describes Ecogame as using “simulation and early machine learning techniques,” it also states that the system was hand-coded, based on sets of rules, and not trained on datasets of existing images. By current technical standards, it is more accurately described as a rule-based adaptive simulation than as modern machine learning (Mason, 9 Aug 2025).

The term is also documentary unstable in the arXiv record. The record for “Serious Game for Human Environmental Consciousness Education in Residents Daily Life” does not contain the paper itself, and the supplied notice explicitly states that there is no available basis to explain how that proposed serious game functions as an ecogame (Du, 2015). Likewise, the supplied record for “Eco-Evolutionary Dynamics of Bimatrix Games” is described as a generic Elsevier manuscript template containing placeholder text rather than a substantive paper about eco-evolutionary dynamics (Shu et al., 2022). This suggests that any encyclopedic treatment of Ecogame must distinguish between a historically specific artwork, a generic family of sustainability-oriented serious games, and a technical class of eco-evolutionary models, while also recognizing that not every record indexed under the term is evidentially usable.

Taken together, these literatures suggest that the persistent invariant of the ecogame idea is not a single mechanic or domain but a feedback architecture: action changes an environmental or socio-technical state, and that altered state changes the meaning, payoff, or consequence of future action.