Economics in the Age of AI: Are Bot-Run Game Economies the Real Deal?

Edward A Thomson
9 min readDec 13, 2023


tl;dr: A game economy is a real economy when there is real money, bots or not.

Recently, an intriguing idea sparked my curiosity, leading me to ponder its validity and explore its deeper implications. The concept, which I’ve come to believe holds merit, is as follows:

Conjecture: A game economy entirely run by bots can be a real economy.

This raises immediate questions: Can a game controlled by bots retain its ‘game’ essence? Furthermore, the use of ‘real’ in this context invites scrutiny. However, I argue that the involvement of real money lends tangible authenticity to this virtual economy.

I envision a direct connection to the world’s actual economy — specifically, the possibility of trading in-game items or tokens for real-world currency (e.g. dollars). This scenario is not merely a fantasy; we ought to see this with the evolving capabilities of AI-powered bots.

My approach to unravelling this idea has been inductive, analyzing the concept step by step. Through this blog, I aim to present a compelling argument supporting the reality of such an economy, driven by the unseen but powerful hands of artificial intelligence.

What distinguishes a game from a simulation?

What exactly defines a game? While there’s an informative Wikipedia page on this topic, let’s delve into some edge cases that are particularly relevant to our conjecture.

Take a chessboard, along with its rules. Most would agree it constitutes a game. But let’s be a bit pedantic: Is a chessboard without players still a game? Does the essence of chess as a game require not just the board and rules but also active players? Otherwise, couldn’t the board be considered mere art?

Consider a computer game, particularly an offline one. We refer to it as a game even when not actively playing. This label remains even if, hypothetically, no one in the world is playing it at the moment.

Now, let’s think about idle games, which typically require minimal human interaction. For example, ‘Melvor Idle,’ an idle game based on Runescape, involves some player input but emphasizes automation. Players can initiate actions that continue autonomously for a period. Despite this minimal interaction, it’s still regarded as a game, driven by the player’s desire to progress their character. The line between a game and a simulation seems to blur here, hinging on human engagement rather than constant interaction.

An in-game screenshot from Melvor Idle (Source: Nexus gg)

Imagine a newly launched MMO devoid of human players. As with our previous examples, we’d still categorize it as a game. But what if this empty MMO were populated entirely by bots? While resembling a simulation used for research, its original intent was to be a game. This is akin to having two bots play chess — a simulation of a game, but does the label of ‘simulation’ or ‘game’ truly matter here?

If the game lacks any form of human intervention, it leans more towards being a simulation. This distinction sets it apart from an idle game. In idle games, despite minimal interaction, there remains a degree of human involvement, be it initial setup, periodic decisions, but not merely the act of observing progress.

A bot-populated MMO, devoid of any human input, transitions from a game to a simulation. The absence of human engagement or decision-making is what shifts its identity, highlighting the nuanced differences between a game designed for human interaction, however minimal, and a simulation running autonomously.

Sir Pugger monitors bot activity in Old School RuneScape (Youtube)

This leads to a philosophical quandary: does a game powered by advanced AI bots remain a game, or does it transform into something else? The distinction might hinge on continued human interest and the potential for human decision-making. While this question ventures into deeper philosophical territory, it’s an essential consideration for our exploration of AI-driven game economies.

As an addendum, let’s briefly touch upon decentralized gaming, which by its very nature, invites a discussion about autonomy and the role of AI. In such environments, where actions are permissionless, the presence and impact of bots become even more significant. This not only challenges our traditional understanding of gaming economies but also raises questions about the future of these economies as they become increasingly intertwined with real-world financial systems.

The possible prominence of bots in decentralized games is the core factor that inspired my initial conjecture.

Generative Agents: Interactive Simulacra of Human Behavior

There is a piece of research that implemented computational software agents (“AI”), which were designed to simulate believable human behavior, within a game-like world. The agents could perform everyday activities, form opinions, and interact in ways that are dynamically conditioned on their changing experiences and environment.

In this work, we demonstrate generative agents by populating a sandbox environment, reminiscent of The Sims, with twenty-five agents. Users can observe and intervene as agents plan their days, share news, form relationships, and coordinate group activities.

The study involved evaluating these agents to test their believability and ability to produce emergent social behaviors. This is a simulation, but the graphics make it appear like a game. The possibility of believeable human behavior is also interesting. To the best of my knowledge, no trading was involved.

The code is also now open sourced.

Economics: What is real?

In our everyday understanding, we perceive our world and its economy as undeniably real, setting aside any Matrix-like philosophical debates for the moment. The reality of our economy stems from its human actors and their interactions.

Consider the realm of high finance, where complex software and algorithms, employed by quantitative hedge funds, trade commodities and stocks. These algorithms, or ‘bots,’ engage with tangible assets and real companies, making their impact on the economy palpable and significant. The intelligence level of these bots doesn’t diminish their effect on our real-world economy.

Contrast this with games, which traditionally exist as self-contained universes, often disconnected from the real world. The activities within most games don’t directly influence our physical reality, aside from considerations like addiction or mental health. In this context, computer games are a form of entertainment within our real economy, a sector where we invest money. However, the characters and actions within these games typically don’t mirror real-world dynamics.

Now, consider an MMO with its own economy. We label this a virtual economy, and it may exhibit characteristics similar to our real-world economy, particularly when the participants are human. Yet, these virtual economies are generally not considered ‘real’ in the traditional sense.

Futuristic cityscape, composed of skyscrapers made from glowing digital code and interconnected data lines (image generated by Dalle3/ChatGPT)

Here comes the pivotal caveat: the moment there is a tangible link between the game economy and the real world. This occurs when in-game items or currency can be exchanged for real money. The digital assets of the game, though virtual, acquire real value, tethered to actual currencies like the U.S. dollar. At this juncture, shouldn’t we start considering the in-game economy as part of a real economy?

Extending this thought experiment further, let’s replace human players with bots. Imagine these bots are controlled by humans who maintain a vested interest in the game’s outcomes. As their bots succeed, they accumulate items convertible into real currency — essentially, they ‘play to earn.’ The complexity of these bots, whether advanced AI or simple in nature, is superfluous to the conjecture. If the entire economy of the game is bot-driven, but there is significant human interest and a direct link to real money, coupled with its influence on our actual economy, then this AI-driven virtual economy transcends its virtual boundaries and becomes part of a real economy.

Ultimately, this could have an impact on world GDP, regardless of how small it is.

What about a simulation with real money?

If a simulation uses real money, is it a simulation or is it real? The primary purpose of a simulation is to model or replicate certain conditions or behaviors for study, training, prediction, or entertainment. If a simulation uses real money, it blurs the line between a theoretical model and real-world consequences.

Humans may have a vested interest in the outcome of a simulation, but that alone would not be enough necessarily to make a simulation a game. The consensus perception is probably what matters (i.e. “it depends”).

AGI in real and virtual economies

The advent of Artificial General Intelligence (AGI) could usher in an era of more complex and dynamic virtual economies. In these environments, the actions of AGI-driven bots might become indistinguishable from those of human players, further reinforcing the conjecture that a bot-run game economy can indeed be real. However, it’s important to recognize that this level of technological advancement, while conceivable, is likely many years away. This prospect raises a profound question: if AGI agents were to replace many human roles in the real world in the future, would our definition of a ‘real’ economy change? Such a scenario challenges us to reconsider our current understanding of economic systems and the role of human versus artificial agents within them.

A futuristic digital landscape that illustrates a virtual economy, complete with a vibrant marketplace, digital storefronts, holographic displays of stocks and cryptocurrencies, and AI-powered bots interacting within the environment. (image generated by Dalle3/ChatGPT)

Economic Behaviour and Market Dynamics

It is reasonable to anticipate that the dynamics of a virtual economy will closely mirror those observed in real-world markets. Just as in traditional economies, key factors such as supply and demand, market fluctuations, and the process of value creation are likely to be equally relevant and influential in virtual settings.

Initially, there may be a hesitancy to fully entrust bots with market-making responsibilities. This could lead to a scenario where human oversight plays a significant role in managing the market, even as bots handle other gaming tasks.

In the early stages of a game’s existence, particularly while the market is still developing and relatively small, it’s probable that human gamers will be the dominant players. The bots created by these gamers might not possess the sophistication of those used for market-making in hedge funds. While there is potential for overlap between gaming enthusiasts and financial experts, it’s not a certainty, especially not in the initial phases following a game’s release.

However, as the game’s economy grows and matures, reaching a point where the total value becomes substantial, we can expect a shift. Sophisticated market makers, likely with more advanced bots, may start to enter the game’s economy. This influx would mark a significant evolution in how the game’s economy operates, potentially bringing it closer in complexity and operation to real-world financial markets.

Conclusion: Validating the Conjecture

As we draw this exploration to a close, it’s important to revisit and highlight the key factors that substantiate the conjecture that a game economy entirely run by bots can indeed be considered a real economy:

  • Human Involvement and Interest: The presence of human interest and involvement, even if indirect, plays a vital role and ensures that it is a game rather than a simulation. In scenarios where humans control bots or have a vested interest in the outcomes of the bot-driven economy, there’s a clear human element influencing and benefiting from the virtual economy.
  • Real-World Economic Link: This is the tangible connection between the game’s economy and the real world. This is established when in-game items or currency can be exchanged for real-world currency.
  • Impact of bots on the economy: Bots have a significant impact on the economies they engage with, they can influence the game economy’s market dynamics and asset values, and their actions have real-world consequences due to the real-world economic link.
  • Economic Behaviours and Principles: The virtual economy, especially when linked to real currency, often mirrors the behaviours and principles of traditional economies. Supply and demand, market fluctuations, and value creation are as relevant in these virtual settings as they are in conventional markets.

In light of these factors, it becomes evident that a game economy, even one entirely run by bots, can hold the characteristics of a real economy. This conclusion not only broadens our understanding of what constitutes an economy but also underscores the evolving relationship between technology, AI, and economic systems in both the virtual and real worlds.


Some of the ideas have been brewing in my head for a while, but the conjecture came to me as a moment of inspiration (cf. shower thought). The idea of bots dominating games has been something discussed within the Decentralized Gaming Association a number of times, especially in the early days.

I used OpenAI’s ChatGPT as a copy editor for this blog.