Lots of warnings, and then it seems you don't write the actual content, suggest that other people do it, and threaten to remove their examples if they go too far. Hm.

Okay, here are some ideas...

• There are crazy people out there, reacting out of proportion. Triggering them on purpose is evil. But even without targeting any specific one -- if you speak to a sufficiently large crowd, one of them is likely to hear you. "Only one person in thousand could understand it that way" sounds weakly from outside, if your audience is tens of thousands. (Problem is, you usually don't grow audience to tens of thousands by speaking moderately.)
• Even relatively sane people may have obsessions or phobias, which can be triggered.
• Lying, especially when it is difficult to get evidence.
• Providing technical knowledge to evil people who lack technical competence. ("Here is the blueprint for doomsday device. This expensive component can actually be replaced by Raspberry Pi.")
• Providing coordination to evil people who until now have been stopped or slowed down by coordination problems.

Now of course, many of these examples are problematic and can be abused. For example, 4 and 5 depend on your definition of "evil" -- some people will be happy to use this as a label for any disagreement. Example 3 depends on what is truth. Examples 1 and 2 can lead to "heckler's veto". Which means, any of them can... and sooner or later will... be used to suppress harmless but inconvenient speech.

Without examples, this post is almost entirely devoid of content.

The general way to get around the infohazard is to use either historic examples or examples of other cultures and societal contexts. 

You can say that saying "Democracy is good" can cause harm because saying it can motivated people to act politically in totalitarian societies in which they will get punished for it. 

Butterflies famously can cause a lot of harm by flapping their wings as well. That seems to be a way words can cause harm as well.

I think most people can agree that in both of those examples the actions can lead to causal chains that result in harm. The problem is that you play motte and bailey and equate "words can be lead to causal chains that produce harm" with "interlocutors should be more careful with their words". That leads to avoiding any of the cruxes that might come up with "interlocutors should be more careful with their words".

Appendix I

First, a little context. When I began the process of writing this essay, I nearly fell into the trap of using a set of words that were callibrated to be harmful as example of exactly that. Seeking to avoid this pitfall, I put up a set of proverbial traffic cones in a wide area around the issue; I decided that no examples would be safer than actually writing a list designed that way. That was true, but I got stuck in a binary mode of thinking that resulted in my missing the possibility of abstractions available in between those extremes.

After I published the essay, the comments showed me a number of ways in which my writing is... a bit rough in places. In particular, several excellent suggestions regarding places where I had set the cones too widely; where my over-avoidance of examples was limiting my ability to communicate effectively, and making it difficult for others discuss my work. Huge thanks to the people who wrote those comments! They reminded me that these things have names, after all, and that we can easily discuss them without accidentally engaging in them or encouraging others to do so if we take a little care.

So for those who feel like examples make a much stronger argument than theory, let's connect some of the dots. Below is a short list of ways that words, and often only words, can be used to manipulate the circumstances in a way that can result in harm. Most of these examples work by convincing the victim that a destructive action is in their best interest. This is typically accomplished either by controlling the conditions or the perceived conditions for the victim, then offering a specific instruction with a harmful outcome, but that's not the only possible strategy. While props and other physical components can be helpful to those attempting these acts, they are not really necessary, and speech alone is sufficient as a trigger once the stage is set.

##  Short list of processes by which words can cause deliberate harm

• Lying is a pretty central way of causing harm, which is probably why there are so many rules made against it. By convincing someone that the world is other than the way it really is, people can get others to act in some pretty destructive ways.  I'm aware that there has been some discussion in the past as to what constitutes Lying, so I'll define it here as "willfully misrepresenting the facts". That includes cherry-picking as well as alteration, but excludes being accidentally incorrect, and also excludes the listener taking the wrong impression in other ways that were not intended. Those can be harmful too; I'm just setting them out of scope for the word "lie" here.
• Sales tactics can get people to buy and consume things that are (physically, ecologically, economically, socially...) harmful to them and/or others, or things that they do not need or want, causing direct economic harm.
• Threatening need not have any non-verbal component, only perceived credibility, to establish conditions in which someone might perform harmful acts.
• A subset of threatening, Blackmail, can easily ruin lives. This works by manipulating the circumstances such that someone becomes willing to do things that are otherwise not in their best interests, usually by threatening to make some information public that the victim believes would cause them harm unless the victim behaves in a certain way.
• Intimidation overlaps with threatening, but the threat can be either explicit or implied by the setting, some aspect of the delivery of the words, or some perceived power held by the threatener. These constitute necessary conditions for words to do their work. Police interrogations designed to be intimidating produce more false confessions than other methods, resulting in innocent people going to jail.
• Bullying contains necessary verbal components including threats and words designed to make the victim question their own worth. Often physical violence accompanies the words, but it may not. This allows the bully to make demands of the victim under threat of more bullying. Long-term, this process has resulted in suicides.
• Gaslighting is a big pack of lies that are orchestrated to convince a person that they are actively and systematically misperceiving reality. The end result can be that the victim loses touch with reality so badly that they begin making actively destructive or self-destructive decisions.
• Peer pressure takes advantage of the mind's willingness to compare it's situation with that of others, and leverages the fact that rejection causes the brain to react as though there was physical discomfort.
• Horrors have been committed by people "just following orders". Indoctrination of people into organizations that expect that kind of behavior tend to mix and match these and other methods very efficiently. Once accomplished, this renders "orders" (mere words) as high priorities for the subordinate.

To be of LW standard this essay should also steelman the opposing argument and assess it fairly 

Ie that Censorship of words can cause harm.

-