By Megan Ma, CodeX Research Fellow, Stanford Law School.
Vagueness has a bad reputation. But, is vagueness necessarily bad? Take, for example, when asked whether you enjoyed the food at your partner’s dinner party; and say you did not. In answering, you might say ‘well, it was certainly textured.’ Vagueness here is strategic, for perhaps it is better not to say forthrightly your opinion and, instead, offer a nondescription of the meal.
And so, vagueness can be both an asset and a liability. There are evidently circumstances when vagueness is unintentional, or worse, could enable disputes. Yet as humans, the vagueness of our words may be a reflection of our implicit positions. In certain situations, we feel it isn’t the right place to commit, we may be hedging risk, or perhaps, we simply do not know. In these scenarios, the use of vague language, infusing linguistic space, is not only regarded as acceptable, but also highly intentional.
The questions become: is maintaining linguistic space necessary and, if so, when do we need it? On the other hand, what are the points at which we have stretched our words such that it becomes a liability? In the context of contracts, disputes related to word meaning arise from the lack of clarity around the definitions governing the terms and conditions behind clauses. As a result, the open texture and the fluidity of language use warrants further investigation, particularly as we transition to using formal languages to draft and construct legal documents. It follows: how could we assess the stretchiness of our words?
Interestingly, Ryan Johnson recently described how the ‘flexibility’ of legal language is a seminal feature that allows the law to be applied in multiple circumstances. This discussion of ‘flexibility’ is where ambiguity and vagueness should first be distinguished. Consider for example classifying whether a tomato is a fruit or vegetable. There are two choices. A tomato is either a fruit or a vegetable. This is understood as ambiguity, and arguably, the flexible quality of legal language. Though we may not always know the choice to be made, there is frequently an exhaustive list of options and a foreseeable pattern of application.
In contrast, when asked whether a burrito is a sandwich, the choices are less clear. While one may implicitly understand what a sandwich is, there is less certainty around what necessarily qualifies a sandwich. Some linguists have suggested a componential approach; that is, what are the necessary ‘components’ that would make up a sandwich (i.e., bread, tomatoes, lettuce, etc.)? Or, is there a structural aspect (e.g., sandwiches must be layered); that is, should sandwiches be defined by ‘prototypical’ variations of them? Assessing which traits qualify a burrito to be a sandwich demonstrate that, more often than not, the parameters that govern word meaning are not as straightforward as they seem. Occasionally, and especially in the legal space, they can even seem far-fetched.
These same lines of inquiry were famously discussed by legal scholar, H.L.A Hart, with the age-old hypothetical of defining a ‘vehicle.’ Consider the following case in the context of insurance. While you could know that vehicles are covered within a given motor vehicle policy, one might not know what constitutes a vehicle. In 1986, the Utah Supreme Court determined that a man was ‘driving’ under the influence when he commandeered a horse (while heavily inebriated) and went for a joyride. But, would his motor vehicle insurer ever have thought that his horsing around is akin to driving recklessly? Should insurers then consider horses as vehicles? Therefore, to the core of our problem: the law’s language is inherently vague. Presently, what qualifies the terms of a contract are stated, but not sufficiently specified. That is, contractual clauses are defined in theory, but not in practice. Accordingly, contractual terms are divorced from their operational realities.
This suggests that prior to considerations of formalization and making contracts computable, we must first tackle a more fundamental question: the relationship between contractual wording and the exercise of it, specifically, the interaction between vagueness and implicit knowledge. Though ambiguity can itself pose challenges in considerations of computational law, the difficulty lies primarily with vagueness.
Nevertheless, as vagueness carries a negative connotation, we suggest in its place ‘elasticity.’ Diagnosing elasticity, and establishing a metric around it, offers a lens behind the unspoken. This is particularly relevant for contracts, where having a deeper understanding of linguistic framing is helpful for both parties. This is because embedded in the language is the DNA of each party’s implicit position. Tracing these linguistic triggers, mapping them, and being capable of gauging their use is, therefore, significant in building robust contracts.
The idea of ‘elastic language’ was first pioneered by linguist Grace Zhang. She identified four key ‘stretchers’ that are commonly found in language: approximate, scalar, general, and epistemic. Epistemic stretchers are particularly interesting. They attempt to characterize the attitude behind uncertainty and lack of commitment. In this way, we may be able to better understand the reason behind elasticity, to determine whether it is indeed strategic and/or intentional, or merely a consequence of conserving explanation.
Understanding epistemic stretchers could, therefore, act as a marker for contractual behaviour, capable of unpacking the relationship between clauses. For example, are there hierarchies amongst clauses? Moreover, what is the intention behind verbose clauses with multiple nested provisions? The answer to these questions would suggest that elasticity is relative depending on the relational structures of contractual terms. A clause may appear highly elastic, but if its general effect on the contractual outcomes is low, it may not necessarily raise any flags.
As a result, further research into elasticity, and linguistic complexity, has the potential to capture insights from legacy contracts as well as enable current and future drafting processes for computable contracts. In effect, understanding elasticity fosters stronger foundations for future computational work in law and ensures a grounded representation of the law’s current language, as we begin to transition into the law’s next medium of expression.
[ Artificial Lawyer bonus comment: you may wonder what this has got to do with legal tech. The answer is that if we need to embrace vagueness in contracts then this changes how we approach using NLP tools to isolate the meaning of certain passages in a clause, and it also changes how we approach the idea of coded clauses in smart contracts – because if X doesn’t actually always mean X, but could mean X, Y, or Z, then how can you ‘operate’ a contract that translates the meaning of words into code and then drives actions based on those instructions?
Thanks very much to Megan (pictured) for this excellent educational think piece. ]
1. Ryan Johnson, ‘Content design for beta.ada.gov: writing for action and flexibility,’ 18F (July 13, 2022), https://18f.gsa.gov/2022/07/13/content-design-ada/.
2. On componential semantics, see Betty Birner, Language and Meaning (2018).
3. Almond Alliance of California et al. v. Fish and Game Commission et al., Super Ct. No. 34201980003216CUWMGDS. For a general summary, see Zoe Sottile, ‘California bees can legally be fish and have the same protections, a court has ruled,’ CNN (June 6, 2022) https://www.cnn.com/2022/06/06/us/california-bees-fish-court-ruling-scn-trnd/index.html.
4. H.L.A Hart, Positivism and the Separation of Law and Morals, 71 Harv. L. Rev. 593, 608-615 (1958).
5. Grace Q. Zhang, Elastic Language: How and Why We Stretch our Words (2015).