Cornell University Jungle: The Predicament and Hope of Web3 from the Oracle's Perspective

The Blockchain Ecosystem is Not "Open"

Although open blockchains represented by Bitcoin and Ethereum have opened their doors to everyone, their ecosystems remain closed. In other words, while we can freely join these two ecosystems, there is no naturally existing bridge between Bitcoin and Ethereum.

In fact, for any two digital platforms, they are assumed to have no interaction until we artificially introduce a communication channel. This lack of interaction exists not only between platforms but also widely between platforms and reality.

Consider this scenario: two people sign a smart contract on Ethereum, betting on tomorrow's weather. We know very well that this contract can only proceed once the actual weather conditions are revealed in the real world. To facilitate the execution of the contract, we need to bring real-world (off-chain) information onto Ethereum (on-chain). Due to this mutual independence between ecosystems, we believe that blockchain is not yet "open" enough, thus requiring oracles and oracle networks to promote information exchange.

In practice, the prediction market platform Polymarket has achieved considerable success with its oracle system based on the UMA (Universal Market Access) protocol, providing us with an excellent way to aggregate information through this decentralized mechanism. We also see the oracle industry collaborating with traditional financial institutions such as SWIFT and Google Cloud, and we observe practitioners engaging in interesting experiments, such as how to introduce AI agents into oracle networks to automatically collect real-world information, and how to establish standards for interaction between chains to facilitate communication.

Starting with the Oracle Problem in Blockchain

Next, we will first introduce the detailed background of the oracle problem and review the development journey of the oracle industry in addressing this issue.

As mentioned earlier, a single blockchain is an isolated system that does not have the authority to access external data; we need oracles to bring information to the blockchain. It is termed a "problem" because we cannot ensure that this external information is accurate and reliable, especially when we overly rely on centralized oracles. Of course, this is not to belittle centralized oracles; in fact, they perform quite well in certain areas. However, in many scenarios, we still hope for a decentralized system.

From an economic perspective, we are not specifically advocating for any one form. If a decentralized system can help us achieve established goals (such as accuracy, fairness, inclusiveness, etc.), we are naturally willing to accept it. Moreover, based on current observations, our optimal choice is more likely to be a combination of centralized and decentralized systems. Nonetheless, we should focus more on the decentralized structures proposed by the industry and the progress they have made.

It is important to note that decentralized oracle networks still face many issues, such as the possibility of collusion among oracle nodes, and the existence of multiple equilibria with insufficient communication, making it difficult to determine the direction we will take. In practice, we have also witnessed several incidents caused by decentralized oracle networks, such as the bZx oracle attack and the Pyth BTC price anomaly. For this reason, we need to better understand the advantages and limitations of oracle networks.

What is the Oracle Trilemma?

We will introduce a brand new concept: the Oracle Trilemma.

You may have heard of a similar concept called the blockchain trilemma, which states that a public blockchain cannot simultaneously satisfy decentralization, security, and scalability.

We believe this new concept presents a more challenging problem. We know that for blockchains, all information is on-chain; for example, in Bitcoin, all transaction information can be obtained online, so verifying information is not that difficult.

However, for oracle networks, we need to aggregate a large amount of information from outside the blockchain ecosystem, which means we must first be concerned about whether this information can be trusted. Secondly, we need to consider how to aggregate and report information to upload it to the blockchain.

To solve these problems, we need sophisticated mechanism design and information design to provide the correct incentives, which is precisely the advantage of economists in this field.

Specifically, the oracle trilemma refers to the fact that no oracle system can simultaneously satisfy decentralization, truthfulness, and scalability.

We can think of a simple example: suppose there is a basic fact, but no node in the oracle network can perfectly observe this fact; the observation of a single oracle will be affected by noise.

However, when the number of oracles is sufficiently large, we assume that the mean of the observations will be close enough to the true value. After observing their respective signals, each oracle needs to decide what information to report to the system; they can honestly report their observations, even though there may be a deviation from the true value, or they may choose to lie, as only they know their own observations. After receiving reports from all oracle nodes, the system will aggregate all the information to provide the final prediction.

The three characteristics mentioned above are the properties we hope the oracle network can possess in an ideal state. However, let's imagine that our oracle network has already satisfied decentralization and scalability, and a large number of oracle nodes report information to the system, which leads to the marginal contribution of a single node's truthful report being negligible for the entire system. When some oracles face high costs of obtaining information, they will naturally consider free-riding. Due to this concern, the truthfulness of aggregated information is not the only equilibrium state. Similarly, the other two scenarios can also be demonstrated.

Ouyang Shumiao (University of Oxford): When the number of oracle nodes is sufficiently large, according to the law of large numbers, the noise will cancel each other out, ensuring that we can obtain an accurate signal. From this perspective, is it possible to balance the accuracy of information with the scale of the system?

Cong Lin: Yes, we certainly hope to increase the number of nodes to help us cancel out the noise. But the problem here is that not all nodes will necessarily report their observations truthfully, and this is an incentive issue. In a "good" equilibrium, if I know everyone will report truthfully, I will likely tell the truth as well, but excessively high information acquisition costs can easily distort my willingness, thereby lowering the quality of the aggregated result and moving it away from the true value. What we just described as the trilemma is even just a weak form; achieving all three goals simultaneously is difficult. In fact, there is a strong form: achieving any two goals simultaneously is quite challenging. Imagine if our oracle network possesses both decentralization and scalability, it means that a large number of nodes are providing information to the blockchain, but we know that the capacity of the blockchain is limited, constrained by factors such as energy, making it technically very difficult to put all data on-chain. In other words, regardless of the truthfulness of the information, operating such a blockchain is itself a challenging task.

Can We Solve the Problem?

As economists, we will attempt to solve these (strong form) problems in a unique way.

First, architectural innovation can alleviate the contradiction between decentralization and scalability to some extent. We just mentioned that the cost of aggregating all information on-chain is quite high, but if we can allow the oracle network to complete the reporting and aggregation of information off-chain, only uploading the final aggregated information to the blockchain, the cost of this process will be greatly reduced, thereby ensuring scalability. More importantly, the reduction in cost will yield more positive results: more oracle nodes will be willing to participate in the blockchain ecosystem, thus promoting decentralization.

Second, introducing a warning mechanism can simultaneously strengthen truthfulness and scalability. Specifically, when participants in the ecosystem believe that the current aggregated information deviates significantly from the true value, we allow them to issue a warning. Subsequently, a trusted but costly third party can verify or dismiss the alert. If the alert is deemed valid, the participating faulty oracle will be punished, while the party that initiated the warning will be rewarded, and vice versa. This mechanism can cleverly balance costs, risks, and incentives, thereby ensuring the truthfulness of information. However, since we often assume that costly verification mechanisms indeed exist and are difficult to manipulate, reliance on trusted third parties naturally becomes a limitation of this mechanism.

Third, the current incentive designs in blockchain are almost all static, while as economists, we are more accustomed to thinking about dynamic incentive structures. Imagine that when a cryptocurrency winter arrives, the value of tokens decreases, and the incentives they can provide are limited. By introducing appropriate dynamic incentives, we hope that each oracle node can maintain good performance during the "winter," as they understand that the reputation they build at this moment will bring considerable returns in the future. In this way, dynamic mechanism design can address the issues of decentralization and truthfulness.

However, even if we completely solve the oracle trilemma, we still cannot fundamentally eliminate the security risks of blockchain, nor can we guarantee the long-term sustainability of the ecosystem. Blockchain oracles have numerous vulnerabilities at both the technical and economic levels, and we need robust architectural and mechanism designs to address these gaps.

What is the Future of Oracle Networks?

In addition to providing data, oracle networks can also integrate with industries such as finance and trade in diverse ways, and the integration with oracle networks will affect the performance, stability, and interconnectivity of blockchains and protocols.

For example, Professor Cong's recent research analyzed 4,988 DeFi protocols from early 2021 to November 2024 to assess the impact of oracle integration. They found that protocols integrated with decentralized oracle networks experienced a 75% increase in assets and a 43% increase in market capitalization within the first month after integration. This effect was particularly pronounced in protocols that completed integration before launching their applications.

Looking ahead, oracle networks outline a promising direction for academic research and industrial development. Overall, the oracle trilemma remains a core challenge, and future research should prioritize developing recursive incentive mechanisms that ensure truthfulness in scalable systems. Complete decentralization may not be our ultimate goal; a hybrid architecture that integrates centralized and decentralized elements may better balance speed, cost, and security. Additionally, exploring efficient and robust governance models is crucial for solving the oracle trilemma and building a long-term sustainable governance structure.

Chen Long (Luo Han Tang): I greatly appreciate the efforts of economists in combining mechanism design with technology. Today's research on web3 reminds me of the influx of quantitative practitioners to Wall Street in the 1970s and 1980s. However, I believe there is a fundamental difference between that and today's field: the current web3 is more revolutionary, aiming to create a trustless world to replace the current institution-dominated world. I have three questions and a final comment:

First, you just mentioned that you are not an advocate of decentralized systems; you want to ensure that these systems are genuinely effective. Based on your years of research and observation, what real progress do you see in this field beyond DeFi?

Second, can you imagine another useful decentralized world beyond cryptocurrencies? And is such a world something we truly need? If such a world does exist, what should it look like? Or, if the future will be a combination of centralized and decentralized worlds, where do the boundaries lie?

Third, how significant do you think the impact of the oracle network is? Will it lead to explosive (or significantly high) growth at the application level? It seems that decentralized systems ultimately need to integrate into this centralized world; in this view, is the marginal improvement brought by oracles quite limited?

Finally, I have one comment. I believe, as you have shown us, that trustworthy data is still very important, and this is closely related to the institutional background of the current world. I think a trustless world is still quite distant from us, or rather, this decentralized system is ultimately just a supplement to our real world, not a replacement. The areas that will benefit the most will be those that can effectively combine both.

Cong Lin: Your comments have actually answered your questions! First, regarding progress, compared to generative AI, the progress of DeFi and decentralized systems has been quite slow. I believe the most successful application so far is stablecoins, especially in a multipolar world, they have the potential to maintain growth.

Moreover, governments and various institutions do not seem to completely oppose stablecoins; in fact, their attitude towards them is more open than towards central bank digital currencies. Another example is decentralized exchanges (DEXs); although they are not yet significant enough to challenge traditional exchanges, the mechanisms they introduce are quite interesting, and these emerging applications are very active. I also believe that the tokenization of real-world assets will be a big deal, and regardless of whether we apply centralized or decentralized systems, oracle networks are very relevant.

Regarding the second question, I find it difficult to imagine a completely decentralized world, but this is actually quite intuitive for economists. We know that everything is about trade-offs; both centralized and decentralized systems have their respective advantages and disadvantages, and we are likely to arrive at an optimal interior solution. This certainly won't please those blockchain enthusiasts who wish to "decentralize for the sake of decentralization." Such groups do exist, but I am skeptical about their ability to disrupt any market forces. I believe people are ultimately driven by incentives; if centralized systems are helpful, why not keep them? In this sense, what I can imagine is a more virtualized world that grows alongside web3.

Finally, considering that we are looking at a combination of centralized and decentralized systems, I believe the impact of oracle networks is quite significant. Centralized systems have many issues, such as high market concentration and excessive market power held by individual digital platforms. Recently, we have seen JD.com attract significant public attention, expressing great sincerity towards delivery workers. However, from an economic perspective, once a platform grows into a monopoly, we cannot guarantee that it won't turn around and exploit employees, suppliers, and consumers.

I believe that decentralized structures are a way to address this issue. Otherwise, we will continuously discuss antitrust issues and have to break up a large platform, losing the benefits of scale.

In this sense, I believe that if we are in a hybrid model, the role of oracle networks will be very significant, as oracles are key to digitization, which is a fundamental issue. If we are willing to believe in a Hayekian setting, where information is highly localized, using centralized nodes to aggregate information will be very difficult, which provides a reasonable motivation for building a decentralized system.

Chen Long (Luo Han Tang): I believe the right direction is to explore the inefficiencies in various worlds and find a mechanism that can combine their strengths.

Cong Lin: One more point is that when we consider information and trust, we even see that a decentralized trend has emerged even on centralized web2 platforms.

For example, due to the explosion of information, we no longer have trustworthy, centralized sources of information. At this point, how to verify information becomes a problem, which is quite troublesome for the information production and aggregation of the entire society. We can either take the web3 route, providing appropriate incentives for information sources, or we can introduce more regulation and verification of information sources.

If we allow platforms to grow without any restrictions, to be honest, I don't know how to trust this information, which is definitely a big problem for society.

Tang Bo (Hong Kong University of Science and Technology): Among the characteristics you introduced in the oracle trilemma, which do you think is the most crucial? In a hybrid model combining centralization and decentralization, which functions do you think can be delegated to centralized systems? Which functions must remain decentralized?

Secondly, what do you think are the main challenges of integrating AI with oracles? When the two intertwine, will we encounter new systemic risks or ethical dilemmas?

Cong Lin: I'll answer the second question first. I think having an AI agent that can continuously monitor and report information is useful; this automated program can effectively replace some of our work.

Moreover, when we need to stress-test a system, we can use AI agents for simulation, which has advantages over traditional ABM methods. This applies not only to oracle networks but can also be used in traditional economic forms.

For related research, if your paper contains both "AI" and "web3," considering that our field is relatively conservative, some might say, "Ah, you're just chasing trends." This criticism is sometimes not without reason, but in any case, I believe writing an academic paper that connects the two is still challenging.

However, for the industry, I think practitioners are exploring the integration of AI and oracles, although they may be a bit overly optimistic and always eager to find the next growth point.

But this is also why I have always believed that economists can participate in this industry and provide our insights. For example, what constitutes true innovation? What is technically feasible but economically unrealistic? All of this requires our rigorous analysis.

Returning to your first question, I believe our views on decentralization are aligned; the path of "decentralization for the sake of decentralization" is unlikely to work. Based on this, I believe scalability is not particularly difficult; at least we have many technical experiences to draw upon. My personal view is that truthfulness is more challenging and more important.

Whether it is the quality of off-chain information or the mechanism for aggregating information, these challenges need to be addressed with changes to the current mechanisms, and we currently lack sufficient thinking about optimal mechanism design, which presents a huge opportunity for economists. When the dot-com wave arrived, we were 20 years late. Now, facing a new wave of web3 and digital networks, I believe this moment presents an excellent opportunity.

Oracle and Data On-Chain

How do the real world and the blockchain world achieve data interaction?

Topic or theme

ChainCatcher reminds readers to view blockchain rationally, enhance risk awareness, and be cautious of various virtual token issuances and speculations. All content on this site is solely market information or related party opinions, and does not constitute any form of investment advice. If you find sensitive information in the content, please click "Report", and we will handle it promptly.