OddThinking

A blog for odd things and odd thoughts.

Ethics of Puzzle Solving Techniques

When I was a high school student, I entered some National Chemistry Quiz competition – it was basically a school-based multiple-choice exam that took an hour or two to go through.

I took the question booklet home, and my father was interested in testing his own knowledge, so he asked me to read out questions while he prepared a meal. He’d give his answer, I’d compare it with mine and we’d debate who was right.

I came to a question which was structured as a dispute between a couple of students over the predicted results of an experiment. It described the experiment, gave both sides of the debate, and asked you to nominate whether one student was right, the other was right or if they were both wrong.

I started to read the question: “Anthony and Janet are chemistry students and are having an argument about-”

“Stop there!” interrupted my father. “The girl is correct.”

I was a little confused, but checked. “Yes, that’s the same answer I gave.” I had carefully considered the chemistry; my father judged the correct answer purely on the climate of political correctness, completely ignoring the chemistry!

Shaking my head, I continued to the next question: “Anthony and Janet are having another argum-”

“Stop there!” interrupted my father, again. “This time it is the boy’s turn to be correct.”

Sure enough that matched my answer.


Of course, using such techniques is cheating, at least a little bit. In real life, you can’t decide questions of chemistry based on the genders of the chemists. You aren’t learning chemistry here, you are learning exam tricks.

I am finding myself using a similar exam tricks to solve certain (newspaper-style) puzzles.

Abstracting away from the specifics of the puzzle (this technique applies to several of them, including Slitherlinks and Hitori), let me describe the type of scenario:

Suppose the puzzle is to colour in a number of boxes. Each box’s colour is constrained by the colour of some of the other boxes, according to the rules of the puzzle.

Take a particular box, b1 – it needs to be coloured black or white according depending on the colour of box b2. The rules state that if b2 is black, b1 must be white. However, if b2 is white, the rules don’t constrain b1.

I don’t know the colour of b2 yet, so I don’t know the colour of b1. At first blush, it looks like I can’t proceed until I have worked out the colour of b2.

However, here’s the dodgy trick. I know this puzzle is being published by a careful puzzle-setter who has ensured that there is only one solution. If b2 is white, then there are two possible solutions – one with b1 white, the other with b1 black. If that was the case, they wouldn’t publish the puzzle. So the fact that this puzzle is published means b2 is black and b1 is white.

I am afraid to admit I am worried about the ethics of this technique. Am I cheating? This technique is a very powerful optimisation, and yet I feel slightly dirty using it. If I try to solve a real-world problem, I can’t assume there is exactly one answer and that is findable.

I mean, if I am debugging some code, I can’t conclude “the bug can’t be in the third-party component, because if it was, I couldn’t fix it.” Then again, maybe I could ask a pair of students to give their opinion.


Comments

  1. What you’re worried about is metagaming. This is a sticky issue, especially in games such as D&D, where there’s the veil of pretence surrounding the game and strategy. You might know what the right thing to do is (and a lot of power-gamers just go ahead and do it), but you must also consider the motivations of the character.

    As for whether it’s cheating probably depends on exactly what you’re doing. There’s a bunch of games where you’re expected to metagame. There’s a host of problems where you’re given two choices, but in reality you’re meant to choose a third (because the two given are red herrings). You could claim this is “cheating”, because you’re picking a choice outside the two, but it’s the correct answer.

    For other games, this is not the case, but not cheating means you’re just making it harder on yourself, or trying to solve a harder problem. You might get more out of that problem (if it’s worth solving) but if the goal is just the solution then it doesn’t matter how you win.

  2. Hangon, what if b1 isn’t even *in* the answer? I don’t know what sort of puzzles you are trying to solve, but not all of them give a complete solution; for example, you might just need to provide the middle word of a crossword.

    If you’re wrong about b1 being in the answer, you will presumably discover that fact when you work through the rest of the puzzle. But this is exactly equivalent to the guess-and-backtrack technique that so many puzzles require. The point being: if you don’t know that b1 is in the answer, then all you’re doing is guessing, which can’t be cheating prettymuch by definition.

    I guess this doesn’t really answer your question though.

  3. Sunny, the definition of metagame you pointed to said that it was “a prediction of how others will make decisions in a game based on their personality or their previous decisions”

    In Julian’s example, who are the “others”? The puzzle maker?

    I think Julian is more concerned about using the knowledge *that it is a puzzle* in order to solve the puzzle. He’s not using any specific knowledge about anyone else. Subtle difference.

    On the other hand I think his Dad *is* metagaming.

    Two comments from me, no good answers to the fundamental ethics question. It’s a hard one.

    I’m reminded of the scene from one of the Star Trek movies where Kirk describes how he beat the unbeatable computer in his final test. Turns out he hacked into it and changed the rules. Now that really *is* cheating. He should have been expelled instead of getting a “commendation for original thinking” or whatever it was. Starfleet Academy are really letting their standards slip. Or they will be. Or something.

  4. Alastair, Sunny,

    You are both right!

    Sunny linked to Wikipedia’s Metagame entry, which uses a definition that, as Alastair points out, is inconsistent with my behaviour.

    However, Wikipedia has a separate entry on Metagaming, which uses a different definition:

    broad term usually used to define any strategy, action or method used in a game which transcends a prescribed ruleset, uses external factors to affect the game, or goes beyond the supposed limits or environment set by the game.

    I think that is an appropriate term here.

  5. I just proposed a merge of these two Wikipedia pages.

    But which definition should win? I like the definition which appeals to specific knowledge about other players. This information is not necessarily available to all players of the game and hence deserves to be treated with deserved suspicion.

    ObSimpsons reference:

    Lisa and Bart are playing rock, paper, scissors.
    Lisa (voiceover): Poor predictable Bart – always picks rock.
    Bart (voiceover): Good old rock – nothing beats that!

    The other definition of metagaming is way too broad and wishywashy for my taste. For example, *all* non-trivial strategies “transcend” the prescribed ruleset, prettymuch by definition. Consequently this definition doesn’t distinguish between techniques that are ethically dubious and those than aren’t.

  6. I actually think both definitions are correct, and despite the murky waters of “transcending” a rule-set as described by Alastair, I think both should be included in the definition. In order to see the difference, imagine a crypto system. If you break it using things like known-text, or some other well known way of cracking the key, you’re breaking the key, but not the system. The solution you end up with will have different degrees of usefulness. If you break the entire crypto system then you can break into anything, anywhere. If you use man-in-the-middle, then you can break specific keys. If you use known text then you can break keys provided you know specific text. If you use ways to get at the session key then you only have the session key.

    Point is, these are all solutions, and in the end your definition of “solution” could be the minimum one or the maximum one, but the one that solves the problem being put with a minimum of implied assumptions is supposed to be “correct”. With known-plaintext, for example, you’re meta-gaming because you know some text which ideally you wouldn’t know. Different levels of “knowing” imply different levels of meta-gaming, and in fact the puzzle might even require a certain level of meta-gaming in order to be solved at all, but I think the definition of meta-gaming should be inclusive of all of these different levels.

  7. “Cheating” is a human concept. Whether you are breaking the rules depends on what the rules are. So the answer depends on what the activity was supposed to do.

    If you are demonstrating scientific ability, you should examine the question data. If you are demonstrating your ability to outsmart examiners, feel free to game the questions.

    If you are trying to exercise whatever mental skill Hitori exercises, keep to the standard Hitori techniques. If you just want the answer, by all means look it up. If you are trying to exercise your brain generally, coming up with new ways to solve the puzzle are excellent; but if you render it trivial you need to move on.

    Assuming there’s a solution to a Sudoku puzzle, and then guessing, is generally cheating. But writing a computer program to solve it is a valuable mental exercise.

  8. I deal with slitherlink/etc. problems of that nature by changing the goal. I’m just just trying to establish a solution, but to prove uniqueness.

    On the other hand, this can give me a hint that I *know* this route is fruitless, and use that fact to help my search for evidence that this route will fail.

Leave a comment

You must be logged in to post a comment.

Web Mentions

  1. OddThinking » Meta-Gaming and Hitori

  2. Puzzled by Software Development