Self-Driving Contracts.

Author:Casey, Anthony J.
  1. INTRODUCTION II. THE CONCEPT OF THE SELF-DRIVING CONTRACT A. Contractual Micro-Directives: The Building Blocks of Self-Driving Contracts B. The Technology Behind Contractual Micro-directives 1. Predictive Technology and Drafting Costs 2. Communication Technology and Performance Costs 3. Monitoring Technology and Enforcement Costs III. IMPLICATIONS FOR THE LANGUAGE AND THEORY OF CONTRACTS A. The Language of Contract Theory in a World of Self-driving Contracts B. Implications for Contract Formation: Lack of Assent, Mistake, and Indefiniteness 1. Lack of Mutual Assent 2. Mutual Mistake 3. Indefiniteness and Agreements to Agree C. Implications for the Time of Performance: Renegotiation, Breach, and Excuse 1. The Reduced Role of Renegotiation 2. The Reduced Role of Breach 3. The Reduced Role of Excuse 4. Dealing with Idiosyncratic Preferences 5. Correcting Machine Errors IV. PRACTICAL IMPLICATIONS OF SELF-DRIVING CONTRACTS: MARKET AND DEVELOPMENT A. Contractual Micro-directives Created by Parties to a Contract B. Contractual Micro-directives Created by Private Third Parties C. Contractual Micro-directives Developed or Governed by the State D. How Should the Law Deal with the Provider Question? V. CONCLUSION I. INTRODUCTION

    This Article explores the theoretical implications of self-driving contracts. We define a self-driving contract as a contract that writes its own terms or fills its own gaps. To be more precise, a self-driving contract has three key features. It is an agreement where (1) the parties set only broad ex ante objectives; but (2) the contract uses machine-driven analytics and artificial intelligence to translate the general ex ante objective into a specific term or directive at the time of performance; where (3) those terms are based on information gathered after the parties execute the initial agreement. (1) We draw an analogy to self-driving or autonomous cars. Just as a passenger in a self-driving car relies on the car to determine the optimal means (direction, speed, lane choice) to travel between two locations and to update its determination to account for real-time contingencies (traffic, weather, construction), the parties to a self-driving contract agree to a shared goal and trust in the contract to direct them on precisely how to achieve that goal in light of real-time contingencies.

    Stated in the abstract, this idea may sound like radical science fiction. But the first generation of these contracts already exists, primarily in the form of self-pricing contracts. The most familiar example can be found in the auto-insurance industry, where parties agree to price terms that adjust automatically based on computer-driven analytics. (2) Similar pricing terms can be found in dental insurance, in short-term rental agreements, (3) and in transportation services. (4)

    As technologies that allow for predictive accuracy and ubiquitous monitoring and communication advance, self-driving contracts will proliferate. Predictive technologies will provide increased information that allows parties to more precisely choose the actions that will benefit their mutual interests. Parties will have the ability to predict with high confidence that, given scenario X, action Y is the optimal course of action to achieve their agreed upon goal. Monitoring technologies will provide access to the information necessary to determine whether scenario X has in fact occurred. And communication technologies will transmit the directives resulting from the analysis to the relevant actors. (5) These technologies pave the way for self-driving contracts. (6)

    The emergence of this form of contracting will have a significant impact on the way we think and talk about contracts. First, it will reveal that the existing language of contract theory is deficient. Self-driving contracts blur the distinctions between rules and standards, between ex ante agreements and ex post dispute resolutions, and even between complete and incomplete contracts. For example, while some will view self-driving and related contracts as hopelessly incomplete, the opposite view--that they are among the most complete contracts--equally consistent with existing literature. This indeterminacy in the language of contract theory and law has led scholars (and will lead courts) to struggle in interpreting and enforcing these contracts. (7) Notions of assent, definiteness, agreements to agree, unconscionability, mutual mistake, renegotiation, and even efficient breach cannot be cleanly transported to the world of self-driving contracts. Analyzing these questions through the language of self-driving contracts may provide important clarity that is currently missing.

    Similarly, some recent proposals for resolving contractual ambiguities (such as using survey technologies) can be understood more clearly once one recognizes that they are just particular forms of self-driving contracts. (8) Thinking about those proposals through the lens and language of self-driving contacts will reveal their costs and benefits--and how the law should treat them--in ways that traditional contract theory does not.

    Most importantly, we identify the questions of who provides the self-driving contract and how to govern the providers as the central challenges for the law. In that context, we will explore first-party providers, third-party providers, and government providers. We draw distinctions between how the law views these alternative providers in sophisticated-party transactions, on the one hand, and consumer transactions, on the other.

    This Article, thus, has two primary aims. First, we begin developing a new language of contract theory, so scholars and lawyers can fully grasp what is at stake when disputes about self-driving contracts arise. Second, we explore some fundamental implications of the emergence of self-driving contracts. (9)

    We proceed in three parts. In Part II, we describe self-driving contracts and the technology behind them. In Part III, we explore the challenges that self-driving contracts pose for the language and theory of contract law. We briefly lay out some doctrinal and policy implications related to those challenges.

    In Part IV, we show that the most important challenge for the law of self-driving contracts relates to the markets that will provide these self-driving contracts. We discuss the likely rise of a market for third-party vendors providing certified computer code to govern contractual relationships. We predict that arbitration services and insurance providers will evolve and likely fill the space as the providers of self-driving contracts. Other platform providers like Uber or Airbnb are also likely candidates to fill this market. Here, we also discuss the important distinctions between the markets for self-driving contracts in consumer transactions, on the one hand, and for self-driving contracts in sophisticated business and finance transactions on the other hand.

    Before we proceed, a quick word on what this Article is not. This Article does not address self-enforcing contracts and similar technologies like blockchain. (10) These contracts--often referred to as "smart contracts"--innovate on a different plane than self-driving contracts. They contain automated provisions that allow for enforcement of specific provisions without resort to adjudication. In this way, they are extensions of escrow technologies like the letter of credit. The innovation, however, is that the self-executing contracts can trigger automated enforcement without a third-party intermediary--that is, without the escrow agent or the bank which issues the letter of credit. (11)

    Self-driving contracts are a completely different category of contracts. Self-driving contracts require judicial enforcement. (12) But they automate the creation and interpretation of the terms that are to be enforced. The key is that the contract uses technologies like artificial intelligence and learning algorithms to update the terms governing the parties' relationships based on contingencies as they arise. (13)


    The self-driving contract, in its pure form, is a simple concept. Parties agree to an outcome they want to achieve and rely on machine-assisted analytics to direct them toward that outcome. The parties might agree to achieve a given outcome, X, and divide the surplus. In the extreme, there are only two terms to which the parties need to agree: (1) the desired outcome and (2) how to divide the surplus. Later, when it comes time to perform, the parties can rely on the machine to set all other terms of performance. The machine gathers information (14) about the present state of the world and makes predictions about what actions the parties should take to best achieve outcome X and then directs the parties to take those actions. It also identifies the surplus that exists and directs the parties to divide that surplus according to their agreed objectives. To do these things, the machine monitors both the external world and the parties' behavior. As facts in the world change, the machine updates its directives, and breach is defined as failure to comply with those directives.

    While fully self-driving contracts do not currently exist, early partial forms are emerging in flexible-pricing arrangements. The auto-insurance example is the most salient. A driver can install a monitoring device that gathers information on the driver's skill and habits, and the pricing terms of the contract automatically update. (15) The essence of this transaction is one where the parties have agreed up front about the division of surplus, but certain facts about the driver were unknown to them at the time of agreement. (16) As a result, they do not know what the surplus will be or how to maximize it. After the information is gathered, the price changes. That price change has two effects. First, it divides the surplus that is now known to exist. Second, it changes the...

To continue reading