Good advice is a useful explanation about how the world works. It’s like a how-to video that helps you get an outcome you want.
Master Yoda advises Luke because he wants the young apprentice to succeed. Yoda’s advice takes the form of a set of words and phrases that sample his worldview, like pixels in a digital photo sampling a place in the real world. Luke hears these word-samples and uses them to reconstruct what Yoda sees.
Words are lossy approximations though. The bandwidth of language is lower than the bandwidth of perception. Sampling with words is like taking a photo with an old camera that can see only a small number of colors. Any set of words Yoda chooses will invariably fail to capture the rich nuance of how he perceives the world. Luke is left with gaps between what is said and what is meant. He’ll need to fill in these gaps before he can use Yoda’s advice.
More gaps appears when the communication process is constrained. Even if word-samples could capture millions of colors of nuance, to cover the entirety of Yoda’s worldview would still require a lot of words. It’s impractical to squeeze all those words into a single session of advice. There might not be enough time for a long conversation or there might be a word limit on the messaging platform. This forces valuable but lower-order bits to be left out. Luke must deal with gaps opened up by what wasn’t said.
To benefit from Yoda’s wisdom, there are two steps Luke must carry out.
If Luke is asked whether Yoda has good advice, his answer centers around step 2. It’s the part of the process where he spends significant mental effort in a deliberate way. In contrast, step 1 is automatic. Luke autocompletes Yoda’s word-samples in the background without realizing it. But just because the process is effortless doesn’t mean it’s working as intended.
When step 1 is run on a digital photo, the reconstruction process is reliable with little room for error. A high-resolution camera does a great job of seamlessly sampling an actual scene in the real world. With megapixel samples, the gaps are negligible. There isn’t much to fill in to imagine what it was really like where the picture was taken. This is not the case with advice.
When Yoda’s words can only sparsely sample what he’s trying to explain, the “photo” he gives Luke will be low-res and have patches of empty space. All this uncertainty allows the advice to be interpreted in many different ways. Luke’s mind hates this ambiguity and suppresses it by filling in the gaps to come up with the most coherent explanation possible, even if it’s not the one that Yoda meant. To maximize coherence, gaps are autocompleted with whatever is most recent, sensational and legible. Luke projects the fragments of his experience that are most accessible into the gaps.
When Yoda is around to notice, this gets fixed. Yoda cares about Luke and has a strong interest in seeing him win. When Luke gets the advice wrong in step 1, Yoda will dig in and spend time to resolve the misunderstanding, even if it takes a lot of back and forth. The more time the two of them spend together, the more context they’ll share. Luke will be better able to read between the lines and fill in the gaps correctly by himself the next time.
Errors in step 1 can be avoided altogether If Luke has prior personal experience with the domain. The coherent explanation that Luke constructs is likely to be the one Yoda had in mind if Luke has enough real-world context to fill in the gaps correctly. Even if Luke gets only a degraded photo from Yoda, he’ll be able to recognize the scene in the photo if he’s been there before.
But when there’s no in-person guidance and it’s an unfamiliar domain, Luke can get in trouble. If the advice is from a podcast interview or social media post, there’s no mechanism to correct the errors from step 1. Since the step is carried out subliminally, Luke is made to think that he’s just a passive impartial receiver of the explanation, rather than its co-creator. It can feel like the worldview that was “given” to him is absolutely the one Yoda was trying to share, even if it wasn’t.
“It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.” — Mark Twain
Luke isn’t going to know that he was hallucinating in step 1 until the explanation is tested against the real world, if that ever happens. Till then he’ll walk around self-assured that his understanding is correct. If the opinion is particularly coherent and strongly held, he may even go on to confidently advise others.
Alas, even if Luke wins in the real world using his explanation, it might still not be a useful explanation.
A useful explanation tells you what to do in a particular situation to get an outcome you want. Most useful explanations are learned from others, but Luke can natively generate new ones when he experiences the situation in person. Luke’s experience of the real world takes the form of a set of sensory inputs that sample the world, like the words that sample Yoda’s worldview.
“Your personal experiences make up maybe 0.00000001% of what’s happened in the world but maybe 80% of how you think the world works.” — Morgan Housel
Since the bandwidth of perception is far lower than the bandwidth of reality, gaps appear between what is perceived and what is there. The gaps from sensory samples can be worse than the ones from Yoda’s advice since Luke might not be seeing, hearing or feeling the actual reason why something is happening. There’s no wise elder to prioritize which aspects of the situation get sensory coverage, like Yoda did when giving advice. Luke alone must figure out where to focus his attention to generate the right lesson. If it’s an unfamiliar domain and he misses the highest-order bits, he’ll be left with with some formidable gaps.
Luke fills in the gaps to explain what he’s experiencing in the same way he filled in gaps with Yoda’s advice. Here again Luke will be unaware of the errors he’s exposed to step 1 when generating a model of what’s happening. Even when his explanation is way off, it can seem to him like he’s simply observing the world that’s right there in front of him instead of something that’s manufactured by his mind. It’s like he’s wearing a giga-resolution VR headset that can’t come off.
It’s obvious that an explanation isn’t going to work unless it’s applied in a target situation that matches the one where it originated. To correctly apply a particular explanation, it’s crucial to see both the source and target situations for what they are. But getting an accurate reading of any situation is really hard if there are significant gaps that need to be filled in. It will be quite confusing and trippy if the situation produced by Luke’s headset is very different from the real one outside. Explanations will work sometimes but fail at other times with no apparent reason. It can seem like the gods are being capricious and unfair but actually it’s just a technical issue with the headset.
When an explanation is consistently useful to Yoda and gets packaged as advice, a new type of target situation mismatch can occur. Yoda wants Luke to win. To get this outcome, Yoda needs to know how to pick explanations from his collection that will be useful to Luke and which word-samples to prioritize. This is a meta-explanation that Yoda will apply when he’s advising Luke. To apply it successfully, Yoda must accurately understand the meta-explanation’s target situation, which includes Luke’s mental state, experiences, edges and glitches. As before, gaps will emerge.
Like Luke, Yoda is wearing a headset that synthesizes the world he sees. When making sense of Luke’s context, the headset projects Yoda’s own experiences and abilities into the gaps since this is the version of Luke that is most coherent. Seeing that it worked in his own circumstance, Yoda tells Luke that he should build a personal brand, learn how to talk in code and move to Coruscant Valley. This explanation isn’t useful to Luke since it’s incompatible with his edges and glitches; he really struggles to code.
Ideally Yoda would have dug deeper to really understand Luke and ensure there was a match with the target situation. If this wasn’t feasible, it would’ve been better for Yoda to ignore his headset’s version of Luke and keep his advice abstract and context-independent, rather than create false precision to suppress the ambiguity. He can also choose to say nothing at all. Since Luke will use Yoda’s advice as a foundation to make sense of subsequent explanations, Yoda should aim to first, do no harm.
Even if Yoda showed restraint though, there’s no guarantee that Luke would also have been ok to leave things hanging. Luke seeks the comfort of being told what to do and how to win. It’s unsatisfying to get a vague explanation that pushes him to do more of the heavy lifting and fill in the gaps so that it’s useful. If Yoda won’t provide the instruction, Luke will most likely look for another advisor who can give him the certainty he craves; an explanation that’s concrete and coherent, even if it’s useless. If Luke isn’t aware of how his advisor’s mind, and his own, can fail him, hearing a piece of advice can lead to worse outcomes than if he had never encountered it at all.
“It is better to be vaguely right than exactly wrong” — Carveth Read
We’re now at the end of this note, so you’ve already had to deal with the many gaps I left. There were some about you that I had to fill in as well. I hope we touched only the gaps within our circle of competence. If we inadvertently filled in the others, I wish us the courage to form no opinion, to ignore what we think we clearly see and to embrace vagueness and ambiguity.