If you’re answering a multiple-choice question and you have second doubts about your answer, you should always switch. In fact, there is 70 years of research backing up this strategy: the majority of answer changes are from incorrect to correct, and most people who change their answers usually improve their test scores – regardless of test content or environment. Like the case of the Monty Hall problem, you should always switch your answer if given the chance (although in the case of the Monty Hall problem, one of the answers you didn’t choose is revealed to be incorrect before you’re allowed to switch). The only reason that we don’t think this is the case, despite it being demonstrably true, is because it’s counterintuitive.
In the case of MCQs, when we make a change our answer and then are wrong, it seems like an easily-avoidable error, causing frustration and making this type of error more available in memory, making its instances seem more frequent than times when changing your answer results in being correct – known as the first instinct fallacy. In other words, experience indicates that answer-changing is a poor strategy, but experience is misleading. This happens all the time – for example, when choosing a different checkout at a supermarket, whenever your original queue speeds up relative to your new queue, you feel frustrated at your decision and are therefore more likely to remember not to change queues in future, as this is more likely to slow you down.
Escalating commitment (also known as commitment bias) holds that when you choose a course of action, for example an answer to a multiple-choice question, prior investment in a decision (including thinking about that answer) makes us far more likely to increase investment in that decision. James Surowiecki argues that groups are far better at making decisions than individuals, as the aggregation of that group’s individual answers tend to be quite accurate. For this reason, on Who Wants to be a Millionaire, ‘ask the audience’ produces the correct answer 91% of the time, whereas it’s 65% for ‘phone a friend’. (N.B. this doesn’t always work.)
As an aside, when combined with a risk-reward system (such as in The Weakest Link), loss aversion is a massive problem: whenever there is a potential of ending up with less as the result of an action, even when taking the risk means potentially ending up with much, much more, our fear of having nothing overrides the logical choice (go for broke).
What this all really illustrates is that by simply telling students that they should switch their answers if they’re in doubt, we’re also saying that, just because they had one thought first and then changed their mind, doesn’t mean that those latter thoughts are more likely to be wrong – in fact it’s the opposite; in other words, you shouldn’t be afraid of changing lanes in traffic, because if it looks faster, it probably will be – despite what experience tells you.
So, to give educators a more accurate picture of what students know and don’t know, as well as to maximise learning (because realising when you should change your answer, by virtue of being aware of what you know and don’t know, is an example of metacognition, which aids learning*), it might be worth telling students, if they’re having second thoughts about an answer, to switch – or to make them really frustrated while they’re trying to revise.
*For more on the interaction between metacognition and the first instinct fallacy, see Higham & Gerrard (2005).
EDIT: fixed link.