The Argument For Spoilers

post by abramdemski · 2021-05-21T12:23:49.127Z · LW · GW · 24 comments

Contents

  1: the puzzle argument.
  2: the intended experience.
None
24 comments

I'll say it right up-front: this is not an argument against cooperating with people who are trying to avoid spoilers. It's obviously a good idea to avoid spoiling people who don't want to be spoiled, so long as such people are sufficiently prevalent in the population that the minor inconveniences of spoiler warnings etc are predictably going to be appreciated.

Epistemic status: devil's advocate. I've been unsympathetic to spoiler concerns for a long time, for more or less the reasons argued here, but my arguments are far from perfect. Should they be convincing? I don't know. I still occasionally avoid spoilers, but not usually.

For an ideal agent, information should not be harmful. This is violated in many ways in practice, but these often point to irrationalities (including imperfect decision theories, like evidential decision theory). We should avoid harmful information when we know we can't handle it, but I also strive to be able to handle information of all kinds.

As JenniferRM mentions in the comments, avoiding spoilers has a chilling effect on discussions about the intellectual content of media. This means that embedding ideas into fiction can actually inhibit their spread in a weird way.

So, why avoid spoilers?

1: the puzzle argument.

Obviously you don't want a puzzle spoiled, right? It deprives you of the joy of solving it. You can't learn if you don't at least try to solve it yourself before getting the answer, right? Spoilers for movies, TV shows, and books are like this too, to the extent that the story presents puzzles. Even if you're just trying to enjoy yourself, learning and enjoyment often go hand in hand -- the reason your brain rewards you for engaging in play activity is because it satisfies evolutionary heuristics for learning (probably).

But is it true? Do we need to avoid spoilers, solving things ourselves, in order to learn?

First, let's consider an idealized learner.

The perfect Bayesian can learn just as much regardless of what order the evidence is presented in. It's not literally true that the only way to learn to do something is to struggle-fail-revise-retry yourself before getting advice. It's not even necessarily the best way to learn. We learn lots of things by being told. (Would you prefer teachers keep Darwinian evolution secret, waiting for students to invent it themselves? An extreme anti-spoilers mindset would suggest something like this! Indeed, a friend of mine lamented being spoiled on all the big important ideas like evolution, relativity, etc.)

Rather than avoid spoilers, I would prefer to develop the skill of "un-seeing", to be able to inhabit the ignorant mindset and model what mistakes I might make, what wrong paths I'd go down, what strategies would be fruitful vs fruitless. (For example, we could put ourselves in Darwin's shoes, and imagine the way he had to think in order to come up with his theory.) I do think we can develop a skill like this to some extent.

But let's consider less idealized learners.

For machine learning, at least, imitation learning is much more efficient than reinforcement learning. For reinforcement learning, you have to find good paths and then learn them. For imitation learning, you're given the paths; you just have to learn them.

Humans are pretty good imitation learners. Children pick up lots of things from those around them.

The full skill of "un-seeing" is pretty difficult for us -- it isn't easy to inhabit the ignorant mindset once we understand something. However, good teachers like 3blue1brown don't just explain the solution, they explain how you might have come up with it yourself. Essentially, a good teacher does the un-seeing work for you, so that (by imitation learning) you don't just get the solution to the one problem, but are actually a little better at problem solving more generally.

Now, it does seem very helpful for humans to do a little reinforcement learning to cement that knowledge in place: to do it yourself rather than only ever watching other people do it. But (pedantically, for the devil's-advocate argument this post is making) I want to point out that a great way to do this is to teach someone else, making you go through the demonstration in detail for them. So you don't necessarily ever need unspoiled problems to learn on. 

Indeed, the skill of teaching and the skill of unseeing are very closely related.

"But you should at least test your skill on un-spoiled puzzles, even if that's not how you learned" says the one.

Maybe so. But if you really understand something, at the S2 level (not just the S1 level), you can see at a glance that you know how to solve a puzzle, without actually solving it. If you need to test, then at some level, you lack understanding.

(Which is fine! But notice that this is not a necessary thing.)

It's like people have a concept of reinforcement as an important type of learning, but it's making them think that they have to stub their toe at some point, otherwise they'll never learn to avoid furniture properly. You don't necessarily have to personally stub your toe. Trying problems is a pretty good way to test how "fake" your understanding is if you're unsure. But there are other ways to notice this, and other remedies. It seems to me that people put trial-and-error learning up on a bit of a pedestal when thinking about things like this.

2: the intended experience.

Even if we set aside the puzzle thing, the author probably wrote with the un-spoiled experience in mind. Getting spoiled changes your experience. You can only get the un-spoiled experience once; after that, you can get the spoiled experience as many times as you want.

First off, I almost always read/watch things just once, so this argument isn't very persuasive for me. I just have a choice between the spoiled and un-spoiled experience. Which is better?

Some study showed that people enjoy things more when they've been spoiled

“It turns out even halfway through a story, you enjoy a spoiled story more, before you get to that spoiled ending,” said Christenfeld.

But that's not my real argument.

My real argument is that I haven't experienced things being worse due to spoilers. I have had my experience of a book harmed by someone telling me about it, but only when they tell me incorrect or misleading information which makes me anticipate the wrong things or misunderstand. On the other hand, knowing a twist ahead of time and anticipating it, and getting to see just how it unfolds, has been enjoyable.

Don't get me wrong -- I've also enjoyed a really good unexpected twist in a movie. Certainly I couldn't have had that exact experience if I had known more about the movie. But there's this other experience, gleeful anticipation of an upcoming twist, which you can't have without spoilers.

Also, if there's something really good about a story, it's usually worth knowing about -- whether or not I will read the whole story. Since I don't have time to read everything people mention, I'd rather people summarize it for me if it's really interesting, so that we can talk about it right then rather than waiting until I've read it. Recall my earlier argument that enjoyment somewhat tracks learning. To the extent this is the case, total enjoyment should be preserved no matter how you learn the information. So if you can enjoy lots of stories a little bit by getting spoilers for them (and only ever reading/watching a fraction of them), that would be more enjoyment total.

(As with the skilled-teacher argument in the previous section, this enjoyment depends on the storytelling skill of the person who spoils it for you, of course.)

That's my experience. Obviously, things could be different for you. However, many readers may have never really asked themselves the question. Is your experience really ruined by spoilers? Or is it just different? Is it worth the trade-off of the experiences you have to lose out on in order to avoid spoilers? Are you avoiding spoilers because it's good for you, or because it's just what everybody does?

24 comments

Comments sorted by top scores.

comment by lsusr · 2021-05-21T22:37:09.086Z · LW(p) · GW(p)

The more I learn about something the less I care about spoilers. Before I wrote fiction, spoilers bothered me a lot. Now they don't bother me at all. Efficiency outweighs not getting spoiled.

Your post reminds me of the Magicians' Code. The Magicians' Code is a double standard. Magicians usually refuse to reveal the secrets of a trick to the audience. However, we also usually reveal the secrets to fellow magicians upon request. Magicians can appreciate the un-spoiled experience after being spoiled in a way non-magicians usually cannot.

comment by JenniferRM · 2021-05-24T05:27:16.466Z · LW(p) · GW(p)

My central objection is that in speculative fiction there are new and good ideas woven into detailed theories about how the ideas could relate to worlds and people and planning and social processes and so on.

And these ideas are worth talking about sometimes. 

And if you put an idea into writing that people feel like they're consuming for some sort of primarily hedonic appreciation, all of the sudden the idea spreads in conversation through society slower than otherwise

And then because of "people who care about spoilers" there's all this dancing about where people can't just say "I read a cool book at it had a cool idea which was X". Because of the ambient culture, and the sense that talking about fiction in plain words is considered rude by some people, now people have to hem and haw and hedge to figure out what it would be polite to say.

Its crazy. Publishing an idea in a delightful format shouldn't slow the idea's propagation down!

Screw it. For me: spoil away! The faster, and more interestingly, and more relevant to the conversation, the better.

comment by Slider · 2021-05-28T16:55:34.462Z · LW(p) · GW(p)

With pasage of time I also connected that antonym to spoilers in the wantedness sense is "teasers". Trailers are enitely made by disclosing information that is intended to sell/enchance the experience. Notably it isn't even that hard to make a trailer entirely out of footage that appears in the film (in some sense "accurate") that is highly misleading.

Towards the "puzzle element" a trailer conveoying what the movie is about can make a viewer understand why they would want it in their life. However a too throught trailer would leave no reason to watch the actual film.

For the "intended experience" factor, movies often have "official" trailers and there the art team has the chance to set up a different set of expectations. But there is expectation that one should not need to watch the trailers in order to start watching the movie. So most movies work for atleast 2 sets of "baseline knowledge". Some movies even exhibit scenes where they subvert the expectations set up by the trailers, a kind of nod to "trailers always lie" that indeed the trailer did lie.

Even if one were of the opinion that film selection process would be enhanced if every movie came with a representative but short trailer it is clear that a trailer can be made wrong. With trailers specifically the failure to sell the experience can be different from making the experience worse. So atleast one aspects of spoilers is that they "antitease", make it less pressing to engage with the work in question.

comment by JohnGreer · 2021-05-21T15:35:27.481Z · LW(p) · GW(p)

The qualia of enjoyment or not is different for different people. I don't have the gleeful anticipation of an upcoming twist. My main reason for avoiding spoilers is it ruins the feeling of suspense.

This also applies to sporting events as well. If I'm watching a fight where I know the outcome, it's much less enjoyable than if I don't know it. If I know someone wins, I'm not on edge when they're getting punched and kicked by their opponent like I would be if I thought those blows may decide the fight.

comment by Dustin · 2021-05-21T22:19:33.986Z · LW(p) · GW(p)

Could you clarify how argument 2 is different from "I haven't experienced disliking chicken livers for lunch so I argue that everyone should like chicken livers?".

Replies from: abramdemski
comment by abramdemski · 2021-05-22T15:14:34.815Z · LW(p) · GW(p)

I've been in public cafeterias and everyone seems to avoid the chicken liver, but when I ask them about it they generally talk about how good everything else is rather than how bad the chicken liver is, and make the argument that you just can't have something else for lunch if you have chicken liver, and the social culture around it is such that even when chicken liver accidentally gets into some other food even then I don't think people are giving it a real chance because the word for chicken liver is literally also the word for spoiled food so they just describe their experience as "spoiled" and it communicates something true but it also carries this heavy connotation and I've seen myself do the same thing and notice the assumption that chicken liver is bad creeping into my thinking for example when discussing some particular dish everyone is talking about their experience of the part with the fish and they ask me how I reacted when I came to that part and i'm like well mine was chicken-livered and they're like ahhh that's too bad and I'm like yeah but really the only problem here is that my experience was different from theirs so i couldn't be part of the shared experience in the same way but it's so easy for sapien brains to confuse that with it being an objectively worse experience because you feel this little ping of regret and your brain puts a heavy weight on those meaninful interactions with other sapiens even if u consume most lunches in private so the shared experience isn't the biggest part of it and besides if everyone would just give chicken liver a chance this wouldn't even be a thing...

Replies from: Dustin
comment by Dustin · 2021-05-22T15:28:48.045Z · LW(p) · GW(p)

So is it the case that everyone would enjoy everything if they just gave it a chance? If not, how do we distinguish between the potential-likers and the rest?

(FWIW, I don't really care about spoilers. All things being equal don't spoil it for me, but I won't give it a second thought if you do...maybe because I'm good at unseeing.)

comment by Andrew Vlahos (andrew-vlahos) · 2021-05-21T19:55:16.410Z · LW(p) · GW(p)

You missed my main reason for avoiding spoilers. It's not because something is intended a certain way or that I think it would train rationality better to not do something, it's because doing things myself is way more fun than having things done for me. I found trying to figure out how to solve a rubix cube myself to be way more fun than being told would have been. (Or figuring out the villain's plot before the monologue, or whatever).

comment by Nisan · 2021-05-21T21:31:20.755Z · LW(p) · GW(p)

Have you had any success learning the skill of unseeing?

  • Are you able to memorize things by using flashcards backwards (looking at the answer before the prompt) nearly as efficiently as using them the usual way?
  • Are you able to learn a technical concept from worked exercises nearly as well as by trying the exercises before looking at the solutions?
  • Given a set of brainteasers with solutions, can you accurately predict how many of them you would have been able to solve in 5 minutes if you had not seen the solutions?
Replies from: abramdemski
comment by abramdemski · 2021-05-22T15:31:54.010Z · LW(p) · GW(p)

A little? I would boldly guess I'm in the top 1% along this dimension, although I have not tested this in any formal way. I'm generally cautious about whether I have hindsight bias, which in itself probably puts me in the top 25% of the population, but of course this is common on LessWrong. Talking with other rationalists, I get the impression that I am uncommonly good at inhabiting previous mental states. To me it seems like the key thing is trying, but, it could just be some inborn capability or something. It helps a lot when you have more explicit models about how you think about specific things, obviously. But I also think there's a Gendlin's-focusing-type skill where you move around your felt senses into the previous configuration.

I would guess the flashcard test would be pretty bad.

I would guess the worked exercise test would be better.

I would guess brainteasers would be in the middle.

Replies from: Nisan
comment by Nisan · 2021-05-22T20:30:14.045Z · LW(p) · GW(p)

Very interesting. I would guess that to learn in the presence of spoilers, you'd need not only a good model of how you think, but also a way of updating the way you think according to the model's recommendations. And I'd guess this is easiest in domains where your object-level thinking is deliberate rather than intuitive, which would explain why the flashcard task would be hardest for you.

When I read about a new math concept, I eventually get the sense that my understanding of it is "fake", and I get "real" understanding by playing with the concept and getting surprised by its behavior. I assumed the surprise was essential for real understanding, but maybe it's sufficient to track which thoughts are "real" vs. "fake" and replace the latter with the former.

Replies from: abramdemski
comment by abramdemski · 2021-05-24T12:51:12.198Z · LW(p) · GW(p)

Yeah, essentially what I'm arguing (poorly) in the OP is that "surprise is necessary for real understanding" is ... well, not exactly wrong, but creating some cargo-cult behaviors which are not strictly necessary.

Taking the RL vs imitation idea from the OP, it's like people have a concept of RL as an important type of learning, but it's making them think that they have to stub their toe at some point, otherwise they'll never learn to avoid furniture properly. You don't necessarily have to personally stub your toe. Trying problems is a pretty good way to test how "fake" your understanding is if you're unsure. But there are other ways to notice this, and other remedies. It seems to me that people put trial-and-error learning up on a bit of a pedestal when thinking about things like this.

comment by Ericf · 2021-05-21T14:33:36.418Z · LW(p) · GW(p)

Unfortunately, you can't know ahead of time if each piece of art will be better experienced spoiled or unspoiled. So, you have to pay the social costs of remaining spoiler free all the time if you want to ever experience great no-spoiler art. Maybe that cost is too high for you, specifically, but it clearly isn't for some folks.

Replies from: Vaniver
comment by Vaniver · 2021-05-21T14:49:57.203Z · LW(p) · GW(p)

I think it's pretty easy to separate things I've recommended to people as "better spoiled" or "better unspoiled"; so long as my threshold / reason for thinking this is sufficiently similar to abramdemski's, then I should be able to freely spoil for him the art that I think can be spoiled with only minor costs (compared to freely spoiling all art).

Replies from: Ericf
comment by Ericf · 2021-05-21T15:19:08.667Z · LW(p) · GW(p)

Sure, that mitigates the costs, but even knowing "this media is better unspoiled" is just a third level of spoiler (and often is enough to figure out the plot twist early, though notably not in a certain Disney animated feature).

Replies from: Pattern
comment by Pattern · 2021-05-21T18:51:54.892Z · LW(p) · GW(p)

My first guess of what 'better unspoiled' was was 'it's not complicated. The trailer gives too much away. Just go in expecting action, and you will get action. That's it.'


a certain Disney animated feature

What's this?

Replies from: Ericf
comment by Ericf · 2021-05-21T19:03:58.366Z · LW(p) · GW(p)

The three levels of spoiler:

  1. Movie 1, 2, and 3 are good. (No spoilers)
  2. "hey go watch Frozen, but I'm not gonna tell you anything about it" (you can intuit that there is some sort of plot twist surprise.)
  3. Kill Bill is a revenge story where Uma Thurman messily kills her former gang in a variety of fights. (#spoileralert, Bill dies at the end)
Replies from: 9eB1, abramdemski
comment by 9eB1 · 2021-05-23T08:10:38.711Z · LW(p) · GW(p)

Seems you are at least missing one if you think telling someone Bill dies at the end of a movie called “Kill Bill” is your last category.

To get overly analytical, you know it’s a possibility Bill dies. In Sixth Sense you may not even consider the possibility what’s-his-name is dead.

comment by abramdemski · 2021-05-22T15:52:15.103Z · LW(p) · GW(p)

A friend of mine considers "good/not" to be a significant spoiler. Which actually makes sense to me, because having low expectations can make me enjoy things more, and high expectations can make me enjoy things less.

comment by Slider · 2021-05-21T15:17:19.451Z · LW(p) · GW(p)

"unseeing" is pretty hard. I think it is a pretty standard LW meme that if you start with the bottom line what you will mostly be doing is motivated reasoning.

Darwinian evolution is not a spoiler for most purposes. Having just random knowlegde doesn't qualify a bit of information to be spoiler it needs to negatively affect the receivement experience. Most encountering of evolution are not meant to be an aesthetic experience. Rather it is technical information. 

The effect of enjoy spoiled content can have the explanation that media that supports spoiling needs to make sense without the spoiler information yet the spoiler information is valid about it. This makes it deeper than average as it effectively has more content (the down side it thinner unspoiled story for same level of effort). Holograms that contain moultple images in them have more information.

I think a person that wants to be spoiled will benefit if spoilers are marked as such.

Book authors and such are artists and preloading the piece by a different mindset is likely a non-artist trying to do an artists job. I am also interested if you have ever suffered from a book author trying to mislead the reader and found it annoying? Is this signifcantly different if it comes from out-of-authorative sources?

The is the posibility that an additional artist could create a new piece of by deliberately constructing an interpretation guiding mindset. That is if a twist can be made better by prefixing it with story that goes into another direction additional prefixing has the potential to create an additional story. However most spoilers do not try to create additonal art/content but rather just spoil to inform or to discuss.

Replies from: Pattern
comment by Pattern · 2021-05-21T18:58:39.291Z · LW(p) · GW(p)
Book authors and such are artists and preloading the piece by a different mindset is likely a non-artist trying to do an artists job.

If I knew you, then might I be able to better recommend art that you will enjoy?*

Might this further generalize to 'creating a different experience'? An arguer may try to know their audience. But if I know you, should it be any surprise if I can do better - not via an overhaul - but with a small tweak. Customarily, this might be better to put 'afterwards' for an experience after the first time, but still.

*Variation on this: another artist creates an 'alteration' but I recommend (or don't recommend) based on knowledge I have of you?


However most spoilers do not try to create additonal art/content but rather just spoil to inform or to discuss.

Is there an objective way to evaluate this?

Replies from: Slider
comment by Slider · 2021-05-21T20:18:30.595Z · LW(p) · GW(p)

Being able to customise to a single receiver is an argument to be a better artist than someone who needs to do one that speaks to a wider audience. For example watching a movie with friends one can interject inside jokes between the slow parts. Those inside jokes would be just confusing clutter if they were part of the mass-distribution part of the piece. But clearly somebody could do the jokes lame or talk over the "public" parts of the performance.

 

Usually when people are infromed about the spoilery nature of some information they will skirt away from them even if they don't drop the topic. Usually spiling happens when a basic modicum of considering the artistic values is not given. Like it could be a very nebolous question whether a given photograph is part of the art of photography. But police doing passport pictures of people is done for identification purposes. Finding them devoid of art would be proper in atleast in that nobody intentionally put any in. However a stylish potrait might still be specific enough that the subject can be identified and in princple could blow somebodys cover. But the optimization pressure for it to be realistic might not be there (the instrumental value of realism/hyperrealism to artistic end might be very concidental).

comment by Pattern · 2021-05-21T18:49:11.903Z · LW(p) · GW(p)

Short version (marked Section S, for skipping, if you read short then long.):

Don't get me wrong -- I've also enjoyed a really good unexpected twist in a movie. Certainly I couldn't have had that exact experience if I had known more about the movie. But there's this other experience, gleeful anticipation of an upcoming twist, which you can't have without spoilers.

Watch it again.

Would this actually work? Or, have you discovered that, far from being spoiler agnostic (i.e. the perfect Bayesian), you find that the information you have about something, going in the first time has profound effects on your experience, and you do have strong preferences about it?


Long version:

13 upvotes (karma). This seems right for this post.


For an ideal agent, information should not be harmful.

It's not about harm. It's about joy. (Or frustrated surprise.)


The perfect Bayesian can learn just as much regardless of what order the evidence is presented in.

New paradigm just dropped:

Learning as encountering/developing models outside your model/probability space.

Example: You come up with a theory, and cannot think of another way it could possibly be. You test it and find out it's wrong.


Maybe so. But if you really understand something, at the S2 level (not just the S1 level), you can see whether you know how to solve a puzzle, without actually solving it. If you need to test, then at some level, you lack understanding.

Uh, what? That's like saying 'I have a puzzle. But if you can solve it, then you can write a program that can solve it, without you ever seeing the puzzle.'


Some study showed that people enjoy things more when they've been spoiled

Have you compared a series of experiences - on the grounds that, reading it the second time constitutes a 'super spoiler' (provided you remember it)?

What about asking people for their preference? (A similar item: If The Arrival is similar to a book (and was developed based on it in some sense), then what is the preferred read/watch order).)


Even if we set aside the puzzle thing, the author probably wrote with the un-spoiled experience in mind.

Sometimes reading a series in the right 'wrong order' improves things. There's more surprises because of misconceptions. Admittedly, I had this experience with a series that already revealed information in an order that was largely, but not entirely, chronological.


Section S, in it's version one/three form:

(1:)

Don't get me wrong -- I've also enjoyed a really good unexpected twist in a movie. Certainly I couldn't have had that exact experience if I had known more about the movie. But there's this other experience, gleeful anticipation of an upcoming twist, which you can't have without spoilers.

Watch it again. (Whether or not this works might vary. But I do enjoy watching a good movie 'blind'.

(3:)

And certain combinations involving unrelated movies as mentioned offhandedly later: "(If you liked Inception, then find out another movie involves dreamworlds on a train + action** in and out of dreams, you might be way more interested.)".)


Also, if there's something really good about a story, it's usually worth knowing about -- whether or not I will read the whole story. Since I don't have time to read everything people mention, I'd rather people summarize it for me if it's really interesting, so that we can talk about it right then rather than waiting until I've read it.

(Inception might meet this criteria.)

Ah, a Spoiler Trope Wiki! This sounds Awesome! (Although it would also be cool to integrate stuff that can serve as an assessment of 'would you read it anyway' and 'do you care about spoilers'*.) Also separate from mentioning the original, it's really interesting to see different executions of the same or similar ideas. (If you liked Inception, then find out another movie involves dreamworlds on a train + action** in and out of dreams, you might be way more interested.)

*For example, if there's a genre that you avoid like the plague...and as you may have noticed, spoilers might serve as a reason to consume the (unusual(/unusually good)) piece of media.

**Why is it called 'action' instead of 'fighting'?

Replies from: abramdemski
comment by abramdemski · 2021-05-22T15:47:45.888Z · LW(p) · GW(p)

Watch it again.

Oh I totally agree that watching things again is usually boring. This makes sense for the Bayesian superintelligence too. I want to separate this from the spoilers question.

Uh, what? That's like saying 'I have a puzzle. But if you can solve it, then you can write a program that can solve it, without you ever seeing the puzzle.'

For example, if you know how to differentiate (in the calculus sense I mean), you can look at a differentiation problem and know whether you know how to do it vs whether there is some twist you don't know how to deal with. Same with any type of math you know really well.