Sanderson’s First Law of Magic: An author’s ability to solve conflict with magic is DIRECTLY PROPORTIONAL to how well the reader understands said magic.
Belle’s First Law for AI as Magic: A human’s ability to solve challenges with AI is DIRECTLY PROPORTIONAL to their contextual understanding of AI.
REMINDER: Sanderson’s Laws are for authors, Belle’s Laws are for characters (us).
Sanderson’s First Law of Magic: An author’s ability to solve conflict with magic is DIRECTLY PROPORTIONAL to how well the reader understands said magic.
To illustrate this point, Sanderson offers a spectrum that ranges from “hard magic” (magic with firm laws that are clearly explained to the reader) to “soft magic” (magic that primarily serves to “establish a sense of wonder”). His own work leans heavily towards hard magic but Sanderson uses the magic of super heroes to illustrate the concept more broadly. Tolkien, on the other hand, has created a fairly soft magical system. Rowling’s Harry Potter books offer a fine example of a middle ground.
Super heroes tend to be considered their own genre entirely but Sanderson makes a compelling case for considering them a form of “hard magic” fantasy. Super heroes have clear powers and equally clear limitations. When a super hero faces a conflict, they reach into their arsenal of powers. Despite their powers, there is usually some sort of confounding factor that triggers one of the limitations and therefore requires cleverness and wit to fit those powers to the challenge at hand. Conflict is almost always resolved through the use of those powers and yet readers and audiences come away impressed by how these heroes used their powers rather than simply being awed by the powers themselves.
Meanwhile, we never really understand magic in Middle Earth with a few key exceptions (such as the tradeoffs associated with the One Ring). The hobbits and their Fellowship take the long road to Mordor even though it seems plausible that Gandalf could have simply dropped them off at the lava pit with his magic. We never really learn the limits of Gandalf’s magic but the one time the Fellowship tries to solve a problem via Gandalf (the Balrog), the plan backfires spectacularly. This leads to my favorite Sanderson corollary, “magic is always an acceptable device if the result is making the situation worse”. More on this later.
Walking a middle road, Harry Potter exists in a world where each book is quite consistent and yet has some variability over the course of the series (according to Sanderson’s analysis, which I choose to accept as fact without an independent rereading). Rowling uses magic with clearly established rules when the characters are using magic directly for problem solving while also allowing for some fuzzier background magic that keeps the world, um, magical.
There are a plethora of other examples that will immediately spring to mind for any fan of fantasy, all illustrating Sanderson’s First Law. If you’d like to dive deeper on the magic (or writing) side, read Sanderson’s full essay. Unlike most of his books, it is a fast read. Plus he explains his own concepts far more clearly and with many more citations than I can paraphrase.
Belle’s First Law for AI as Magic: A human’s ability to solve challenges with AI is DIRECTLY PROPORTIONAL to their contextual understanding of AI.
Because Sanderson wrote his laws for authors who are building entire worlds and magical systems from scratch, his audience must understand their magic in its entirety. Belle’s law is for characters (us) who live in a world where magic (AI) already exists and are looking to use it.
That is actually really good news for the 99.9+% portion of humanity who has no idea how, and how well, AI actually works because all we really need to know is how AI works within the context of the very specific challenge we are hoping to address.
If all you are looking to do with AI is to leverage the technology as an email reviewer, you will want to know about its propensity for em dashes. You will also want to know about its tendency to fall back on common phrasing rather than your unique voice over time (even when directly instructed to do otherwise) and other related limitations or quirks. Even this incredibly focused application can end poorly if you know only that AI “can” edit your emails rather than assessing how it is likely to do so and considering how that will add or detract value for you.
If, however, you are looking to replace an employee or an entire team with AI you will need to build an equally deep understanding of the AI you intend to use as you did in the context of email editing but for a more extensive scope. Any given employee, let alone a team, has broader responsibilities than editing emails and so leveraging AI in this expanded application requires an expanded understanding of AI capabilities and foibles.
There is a potential pitfall here where an AI user may believe that they do indeed understand AI within a given context but lack the expertise to make a full assessment. I have previously written that AI allows non-experts to deliver on functional requirements without exposing nonfunctional requirements (even in non-technical fields) and this could become an issue even in a domain as focused as email editing.
For example, a user may well know the basic mechanics of typing an email and even possess a strong vocabulary that is bolstered by an intuitive understanding of grammar. That same user may be woefully ignorant of the impact grammar and punctuation can have on tone or how tone must be varied within their particular corporate culture depending on the seniority of the recipient. There are any number of ‘nonfunctional’ requirements for email writing that routinely land humans in hot water.
Of course, if you aren’t trying to solve the problem of landing yourself in hot water, just trying to save yourself the time of typing your way there, this isn’t actually an issue.
The point is that you must be able to accurately articulate the full scope of the challenge you are looking to address if you are going to be able to assess AI’s capabilities within its context. This gets increasingly complicated when those challenges are person, team, or organization sized and yet we humans have been busily developing this type of expertise for thousands of years. We just need to remember to apply it in this context.
AI, like magic, will surprise you. Some of those surprises will be delightful and others will be devastating. The better you know your challenges, the better you can get to know your AI and your choices will be wiser, your surprises fewer as a result.
Neither First Law is truly a new concept. In fact, both are effectively variations on the colloquial Law of Unintended Consequences.
Sanderson’s First Law tells us that readers won’t buy the story unless the characters’ actions can be reconciled with the Law of Unintended Consequences. Plausibility depends on characters being able to avoid unintended consequences because the magic they use follows clear rules and the characters understand those rules OR upon them facing those unintended consequences after all (I told you we would come back to the Balrog).
Belle’s First Law basically says, “you should have reason to believe you are knowledgeable enough to escape the unintended consequences before you add AI into the mix.”
Part of what got me so excited about writing this series was realizing that Belle’s First Law is also a ‘research-based’ expansion of something I have been saying for a few years now: If I know what ‘good’ looks like, I should use AI to get there but if I can’t recognize ‘good’, I should pay someone who does to use the AI for me.
This is still true, however, my original framing left me vulnerable to the unintended consequences in a second hand manner. Seeking an expert in a given domain is wise when you need expert level results, however, Belle’s First Law tells us that we should specifically seek an expert who has invested the time and effort to understand AI within the domain of their expertise. Belle’s First Law improved my own policy and I hope it will do the same for you.
AI is a given at this point. The questions are who, when, and how. Belle’s First Law is the first plank in a foundation for answering those questions.
Check out the rest of the series:
Previous
