It can often be very difficult to determine if a user is genuinely confused or upset versus simply trying to cause problems for laughs or malice. What are the most reliable indicators that a user is intentionally being difficult?
It's often difficult to detect a genuine troll from a confused user. As @AffableGeek stated it's best to give the user the benefit of the doubt at the start and treat them as if they are a new user who simply isn't aware of the rules or how things are done on the forum.
One of the best ways to catch a troll is to watch for trends and to use the tools available in a community. With a combination of the following you can normally catch trolls out:
- User notes: Most communities have some way to keep notes on a user. Moderators can use these to keep track of warnings / notices handed out to users suspected of being trolls. If several threads / messages attract similar notices then you have a pattern of behaviour.
- History: By going through a user's history you can check for a trend or pattern of behaviour. Has the user done this before? Do they make a habit of inflaming comments?
- Private communications: Trolls seek attention, so taking things private denies them their audience. How do they respond to private messages asking them to behave and respect the community rules?
- Penalties: Lots of communities will penalise a person with restrictions and even temporary bans before permanently banning someone. Has this person been subject to these restrictions in the past? Have they responded to them?
- Other moderators: What do your peers think? Discuss things with other moderators and get their feedback. Human discussion is often invaluable in identifying a troll.
I see two fairly reliable, interrelated mechanisms for determining trolls:
- A first time offense is probably confusion
- Repeat offenses from the same user are more likely intentional
User's Attitude when confronted
- If they are apologetic and try to repair what they've done, chances are it was initial confusion
- If they seem excited by the attention and try to escalate the situation further, chances are it is intentional.
Tone is always quite difficult to convey in text-based communications and the short comment system here makes that even more challenging. Nevertheless, the patterns I've found tend to involve repetition, either within the same sequence of exchange or across other messages. It varies a little, but the hard-cases will tend to use more inappropriate means as they escalate. In other words, on a site like this, they start with a comment or two, then they move onto using the answer mechanism.
Either way, I usually assume the first time is just a mistake, but I then watch for repetition after they've been communicated with.
I agree that history is the best place to look for information; just a single interaction is not enough to make a judgement. But unlike the other answers, I don't think that the amount of conflicts is the best thing to look out for.
Trolls are flexible. They adapt to situations, exhibiting different behaviors, just like the average person. They intentionally choose the reaction which will lead to their goal - annoying others. They behave almost like reasonable people, but manipulating the situation towards discord. The better the troll, the more subtle the manipulation. You may even miss it.
Users who cause mayhem frequently without being malicious tend to be rigid in their behavior. They tend to end up in the same kind of drama over and over again. Unlike a troll, they don't want to make the situation difficult. They just can't help it - they have some kind of issue, and when somebody pushes their button, they explode. Unlike a troll (who is playing a mind war against you and is aware of all facets of the situation, enjoying your confusion), their intention is to cooperate, they just don't realize that their actions and words are disruptive and frequently cause the opposite of what they want to achieve.
So, do look for patterns, but the more real patterns you find, the more likely it is that it is somebody who just has trouble fitting in and needs guidance from you. A troll is somebody who is much more slippery, he isn't consistent in his conflict history, may even switch sides completely and argue for something he demonized last year. His dominant pattern is that he maximizes discord, frequently in very creative ways.
For the most part, you'll do the most good by concentrating on individual behaviors and patterns of behavior rather than on what you think the intentions of the user might be.
If an individual message or conversation seems problematic, and if you believe it would be helpful (or mandated by your duties in your role in the community) to intervene, then you should do so, pointing out politely, clearly, and constructively what the problems with the behavior are, leaving out any speculation about the user's motives. If the behavior came from genuine ignorance of or confusion about the relevant community standards, this reaction should be helpful. If not, and you're dealing with a troll, the reaction could still be helpful to other users, and your focus on the behavior rather than the user gives the troll less of a foothold to escalate the conversation into personal attacks.
Similarly, if you see a pattern of behavior from a particular user that is undesirable, call the user on it, again referring only to the pattern of behavior and not to the user's motives. For example
I have noticed many posts from you that question other users' motives. This can be harmful because ... Please refrain from doing this, and instead focus on the issues you may have with the content.
I have noticed that you are trying to pick fights with other users. ...
Like with constructive reaction to an individual post, this sort of reaction may prove helpful to the user or to other users, or it may not, but at least, it focuses on the behavior that everyone can see and discuss rather than on the user's motives, which only that user has access to.
At a certain point, it may become clear that the patterns of misbehavior from a particular user show no signs of abating, despite your interventions, and, in your estimation, the harm likely to be caused should these patterns continue will outweigh the good contributed by this user's participation. What point that is is going to depend on a lot of local factors and your judgement on the scene. At that point, it makes sense to take available actions to limit or stop the user's participation. But note that if you make the estimation based on patterns of behavior rather than your perception of the user's intent, your action will be on much more solid ground.
I have never since that moment in my first year questioned the motive of another member of the Congress or Senate with whom I've disagreed. I've questioned their judgment.
I think that's why I have the respect I have and have been able to work as well as I've been able to have worked in the United States Senate.
Treat suspected trolls as you would treat any other user
My experience in various types of communities makes me say that users suspected of trolling should not get any special treatment. You should judge behavior patterns. Who is a troll? It is a user who intentionally creates a mess to get attention. But judging intent is a bad idea because you can never actually read it.
For example, one LARP community has very complex rules regarding AoE abilities that are really hard to read in LARP, as there is no computer doing calculations for everyone, nor a chunk of time big enough to do it slowly. If someone announces an AoE effect, all other players have to obey what is said. If an AoE user is caught not being able to use the announced ability, he/she suffers very harsh punishment, as it is impossible to check whether the user cheated intentionally or just forgot something. As soon as it is clearly stated in the rules, players understand how things go and act more carefully: potential cheaters don't cheat, other players know the rules well (which is actually rare for complex LARP rules).
Another example is the treatment of problem users on SE. A user gets their chance to learn, but if the chance is missed and the user creates too much mess -- well, you know it. If you assume bad intention and accuse the user of trolling, you are potentially harming someone who is not that bad, just mistaken. If you give some troll a chance to troll for a bit longer, that won't change things too much, because if behavior patterns don't change, the user will soon lose their ability to troll.
In short, again, as you cannot really read anyone's intent, don't pretend that you can.
There is no certain way to detect a troll.
So, what should we do in such situations?
We should remove the term "troll" from our dictionaries, judge actions rather than their doers, and be positive (uncynical) about community members.
In this answer, I mention some points indicating that we should avoid considering any community member a troll.
- The term "troll" is too vague and so highly abusable. It is commonly said that a troll is a person posting in a community only to provoke other community members. Can we judge someone by looking at some of their actions? To clarify what I mean, let us consider this example: Suppose that a new community member objects to some of the community mainstreams and criticizes them. In such a situation it is highly expectable that old community members dislike such criticism and may be disturbed by the member's words. Now, should we call such a member a troll? (If so, then we should call many scientists who have criticized their community mainstreams, such as Galilei, trolls!).
In this Wikipedia article it is stated
"Application of the term troll is subjective. Some readers may characterize a post as trolling, while others may regard the same post as a legitimate contribution to the discussion, even if controversial."
Unfortunately, it seems that many people apply various terms (such as "infidel" and "troll") for calling others to evade being criticized. Such people often abuse this term to silence criticisms against themselves or some community customs.
In a community in which the term "troll" is often used the community members are not motivated to criticize the community customs, so such a community should not be expected to develop successfully.
- Why do we need to judge members themselves rather than their actions? Each community should already determine well-defined norms, and ask all members to behave according to them. Now, if a member violates the written norms (such as the code of conduct or any community rule), the community managers (moderators & administrators) can temporarily suspend the user or remove them from the community if needed. But, if a user does not violate any written norm, but you, as a community manager, dislike their behavior, you can improve your deficient written norms to not allow the members to commit such a behavior; therefore, in this case we cannot call such a member a troll because no one not violating any written norms should be scorned. However, it is recommended that one should not take any severe action against such a user before writing the revised norms out.
- People usually dislike a community member who often questions the community mainstreams or criticizes some actions, so such a member is likely called a troll by others. We should not call any one expressing their opposing opinions a troll. Being criticized is harsh but more helpful than any admiration. As you know well, there are some Q&A communities with a huge number of users, but only a very few number of their members participate in meta issues. Why? Because many people are afraid of being called trolls in case they want to express their novel ideas. If you want your community to be developed successfully, you need to pay its cost, namely being
In other words, any criticism not violating any written norms should be considered acceptable, and in this case no critic should be called a troll.
- As you clearly stated your bewilderment in the OP, it is very difficult (in fact, impossible) to determine whether a community member is a troll or a confused member needing help; it is also clear that the other answers to this OP provide their solutions in doubt. This comes from the fact that trolling is a covert behavior, and as this answer explains, human intent cannot be judged.
Some people introduce some indicators of being troll, like the member's history or their attitude towards some specific actions, but human mind can be too complex to be read by examining overt behaviors. Two persons can behave in the same way such that one of them really needs help and the other just trolls. So, literally, the answer to your question is "there is no certain way to detect a troll".