Rationalist taboo is the name coined by Eliezer Yudkowsky of Less Wrong for a technique which tries to overcome ambiguity and seek clarity in a discussion by restricting your use of language. It specifies that the participants make one or more of the vague terms in a discussion—like "freedom" or "abortion"—a taboo word that temporarily cannot be said: more precise descriptions must instead be used. It is important to note that simply using a synonym does not suffice: the goal of the rationalist taboo is to prompt speakers to carefully define the actual content that might be otherwise disguised by the tabooed word. For example, it is meaningless to argue about whether or not a particular sentiment is "racist," unless both speakers agree on the exact nature of "racism." Making "racist" a taboo word encourages speakers to describe the specific offense entailed, rather than arguing over the validity of a vague label.
The rationalist taboo is one of a variety of techniques applied by the Less Wrong crowd in their attempts to work around perceived shortcomings inherent to both language and human nature. In this case, the target is the imprecision found in frequently-used and poorly-defined words, which allows entire conversations to occur without any real engagement even if both participants have open minds.
Yudkowsky describes the technique as follows:
“”When you find yourself in philosophical difficulties, the first line of defense is not to define your problematic terms, but to see whether you can think without using those terms at all. Or any of their short synonyms. And be careful not to let yourself invent a new word to use instead. Describe outward observables and interior mechanisms; don't use a single handle, whatever that handle may be.
The rationalist taboo comes from the importance of precision in arguments, and also from the need to avoid unproductive arguments from semantics and definitions. Anyone who has ever spent time "discussing" or "debating" on the internet may well recognize the situation where instead of discussing an issue, you actually end up discussing a definition instead. Perhaps Bill Clinton's infamous "It depends on what the meaning of the word 'is' is" was taking this to extremes, but whether it be "democracy," "rationalism," or "reality," it's important to know exactly what words mean and then stick to this convention. The simple fact at play here is that a word is not the same as a thought, and a word itself is quite meaningless without such an agreed upon definition—the most important part here being "agreed upon." Thoughts map onto certain words, but unless both parties have the very same map, their thoughts will not sync up and the discussion can go absolutely nowhere. The rationalist taboo proposes that completely avoiding these words that are the root cause of this mapping error is a technique that can be used to increase the productivity of a discussion.
The point, from a "Yudkowskian" perspective, is to guide people to discussing things in terms of their observational qualities (what we expect to see and experience from them) rather than in terms of just their labels and learned behaviour. It enforces people to visualize details and describe what is really there, often as if you're seeing something for the first time, rather than use a less well defined mental shortcut. It's not too dissimilar to the artistic tip of "drawing what you see".
This isn't just important from the viewpoint of precision, one of the most useful aspects is that it removes deep emotional attachment to certain terms. It is easy to effectively commit emotional blackmail on an opponent by questioning why they wouldn't be for something like rationality, feminism, freedom or democracy—because these words are almost universally considered to be good things that you should agree with. Tabooing these terms frees anyone from the need to unquestioningly agree and focus on what is actually meant. It prevents this emotional blackmail aspect, allowing someone to critically evaluate an idea without having to avoid being labelled as "anti".
 Defining terms
Defining complex terms in advance has long been part of the normal "tool kit" of philosophical and rationalist debating. For example, Scott Clifton, operating under the YouTube name Theoretical Bullshit, made a similar point underlining the importance of this in his 'Treatise on Morality'. Here he took great pains in defining his terms for "good" and "evil": stating that if anyone was to disagree with his definitions it would be okay, and he would just find some other random and arbitrary combination of vowels and consonants to represent his ideas. As if to illustrate how much emotional attachment and preconceived notions of a word can control people, one of Clifton's critics seemed to miss this point when describing the death penalty as not a form of murder—despite Clifton clearly defining the idea he wanted to convey that did define the death penalty as a form of murder. Yudkowsky's formulation of the rationalist taboo goes one step further than this traditional methodology, stating that defining terms in advance isn't a necessary exercise if you can substitute the controversial word with a more precise definition throughout. This avoids the emotional attachment issue that can prevent people from accepting a definition given in advance.
 Critical thinking
Tabooing certain words also acts as a critical thinking exercise for oneself. This is a more subtle application and is only slightly related to the idea of precise communication between people, but it is an extremely important aspect. Often, an argument can end up relying on buzzwords and jargon to support it — Yudkowsky wrote a similar piece discussing the idea of "applause lights," where a term is not really being used to convey anything other than for an audience to applaud it. This can indicate that perhaps even the person proposing the argument hasn't really understood what they mean, or why they mean it, and so might not have the best and most correct reasoning they can be capable of. By removing the vague terms and forcing people to exemplify their ideas, the rationalist taboo comes in useful for crystallizing arguments that would otherwise be too reliant on ambiguous or broad terms to be meaningful.
 Worked example
In the old metaphysical proverb "if a tree falls in the forest, and no one is around to hear it, does it make a sound?", there seems to be a paradox between whether a tree does something, or not. By tabooing the word "sound", clarification can be made in both the question and the responses - and what people infer from the word "sound" is dissolved. People can realise that disagreement of "it does make a sound" and "it doesn't make a sound" are, in fact, illusions caused by this.
The reason tabooing "sound" would work in this case is that the alternative is to form some kind of argumentum ad dictionarium, where people argue based on definitions. I.e., "if you define sound like this, then yes." - "but if you define sound like this, then no." This would lead to a disagreement over a definition, rather than efficiently resolving the conflict by dissolving the primary problem.
This very generic example has a wide range of applications - it can be quite good to bring up in arguments about whether Pluto is a planet. Simply add a taboo on the word "planet" and you quickly realise that pages and pages of internet debate has been generated over, quite literally, nothing.
 It works!
Take the assertion that "prayer works." Although applied more generally to anything that can be said to "work" (such as most alternative medicine), this specific example can have an alarming number of connotations associated with it, and often leads to disagreements between skeptics and believers. A skeptic will ask for the proof that prayer works, while a believer asserts it works on faith and can't be measured by science and proof. So, what if these were removed and rephrased? "Prayer" is an easy one to deal with, as it's simply "appealing to a deity." This doesn't accomplish much but it does serve to differentiate prayer from forms of meditation and self-belief. It would focus our attention on the purported mechanism of action, rather than mere introverted self-belief that could cause prayer to "work" regardless of whether it is based on bullshit.
What about "works"? This is perhaps the key point in the argument. People have a vague notion of what "works" means, but in the context of prayer it may need to be expanded upon. A skeptic would suggest "produces results" — this helps raise the issue of what these results are. They would have to be measurable results, certainly, as results that can't be measured are hardly results at all. Even this might cause conflict; what are "results," what can be measured? So that too can be tabooed to produce a further and a few more levels can be added until we hit a more detailed statement such as "appealing to a deity produces a change that is out of our direct control, that is in line with our expectations that come from what we asked the deity for." This leaves us with a more precise statement, free from the very simplified notion of what "works" means, although it is considerably wordier. It can even be compared with reality. The question is, does a believer come to the same conclusion when they say "prayer works," and can it be reconciled with the precise term generated by tabooing "prayer works" from a skeptical perspective?
 Is it rape?
This slightly "controversial" example could almost certainly benefit from this technique. In recent years many politicians and public figures have argued "is it rape?" as a way of excusing acts, or adding additional qualifiers rather haphazardly to confuddle the issue. Instead, consider the rationalist taboo approach to "rape" as a word itself. Someone then can't excuse some acts by redefining them; and the question "is it rape?" needs to be approached differently.
In this case, the question would be most readily replaced with "was informed consent exchanged?" Here we have something far more concrete to ask; and something far more closely related to whether it is moral or not (in fact, most people would agree this is the question that determines right or wrong). If the answer is "no", then there is no way to really excuse an act based on it being forcible, legitimate, or honest.
Oh, it was forcible consent wasn't exchanged?
 Terms of art
While controversial or flexible definitions in philosophical discussions can benefit greatly from the rationalist taboo, it's not always necessary or even desirable to do in all cases for practical reasons. This is the case with a term of art, an agreed upon definition designed to reduce ambiguity in discussions because it is assumed that all parties hold the relevant information to decode that word with precision; where a party does not often leads to not even wrong circumstances. For example, the second law of thermodynamics has a set definition and a set implication that all relevant scientists agree upon. Substituting the four words "second law of thermodynamics" with anything less than a full textbook of information could lead to increased ambiguity, as the nuances and caveats are removed, and doing this for every mention of the phrase would be just plain inefficient. However, closer to the spirit of the technique, putting a rationalist taboo on certain terms of art can test if someone has actually understood them properly to show that they're not just using the words as uncritical jargon.