Intentions matter. When assessing the merit or moral value of an action, we do not do so solely based on their outcomes, but take intentions into account. For instance, we consider it worse if someone breaks one cup in an attempt to steal cookies than someone who accidentally breaks 1o cups trying to help with the dishes. Only very young children disagree (Piaget, 1932).
Here is the rub: We can only observe actions and their outcomes. Intentions have to be inferred. In everyday life, we have enough data on the people around us (their previous actions, their character, etc.) to make this inference somewhat reliably. As so much rides on a correct inference – the example above shows that the outcome is essentially meaningless in moral terms, unless intention is considered as well – getting it right matters. A lot. Yet, the dedicated social neuroscience circuitry that most of us possess to make these inferences responsibly does critically rely on a rich and subtle dataset of temporal and social context. If this context is missing, this circuitry will still operate, but in a vacuum, making inferences that are no longer reliable (but feel no less compelling to the mind that inhabits that brain).
This is where the modern (social) media landscape comes in. Most of the issues brought before us now are extremely complex and unless we are experts on everything, we have no way of assessing what is really going on, or what the intentions of the actors might be. More often than not, when we encounter someone’s actions online (often mediated by a third party, the “media”), this is also the first time we encounter that person as well. Put differently, we have no temporal context to judge them, nor much in the way of character to go by. But we will judge their actions implicitly or explicitly, and to do so with conviction, and we will also ascribe intentions. As they cannot be grounded in empirical fact – as we don’t know them – something else will be filling in for that. There are plenty of things that can be used in this fashion. Prior experience with other people comes to mind. If one has the experience of being taken advantage of by other people, one might be forgiven for adopting a generally misanthropic view and distrusting people in general – ascribing an intention to cheat to a freshly encountered person. This can be hard to override empirically. If I am convinced that someone is going to cheat me, even them being nice can be construed as a ruse to gain my trust or put me at ease, seemingly consistent with the ascribed intention.
As a matter of fact, it is hard to conceive of a pattern of behaviors that is inconsistent with this perceived intention, which – in this case – is made up without regard to the current situation. This is not an issue in the use case for which the system was designed (or evolved!) to work. Specific brains create a range of specific actions and by observing them over time and in various situations and environments, I can make inferences about the kinds of intentions this particular brain is prone to, which we call character or personality. In this case, everything works out. Brains generate behaviors to further certain intentions, both of which are consistent with a certain character. Once identified, this can be used to predict future behaviors. But in an online environment, the only thing one has to go by are actions. Also, they are often not observed, but reported, and reported by parties that are rarely disinterested. How is one to infer intentions or character? As pointed out above, there are plenty of sources that will allow conscious or unconscious “filling in” of missing information.
One such source is the report of the behavior itself, as it can emphasize or de-emphasize certain aspects of the action, making a particular intention more or less likely. Moreover, ideology is a key aspect of one’s intention-inference machinery, much of which is happening unconsciously and automatically. The problem is that there is very little in the way of “overriding” these conclusions. First, one might not want to override them because they reinforce one’s worldview. Second, there might be no perceived need to do so either, as all possible actions are consistent with one’s inferred character and intention. For instance, if one held the belief that all rich people are evil and only motivated by greed, there is quite literally nothing a billionaire could do to dispel this preconceived notion. If the billionaire gave the money to charity, one could argue that it is not enough or should have been done earlier. One could argue that the billionaire is only doing this to soothe a guilty conscience or to whitewash their legacy (this is not without precedent). One could even go so far as to say that the billionaire shouldn’t have had all that money in the first place, that it was effectively stolen from society and that it is only proper that it is now being returned, no thanks to the billionaire. At no point does the billionaire – despite all that money – have an action at his or her disposal – that could overturn the preconceived notion of the critic. This has happened. What has not happened – and generally does not happen – is for people to adopt a scientific stance and to ask themselves what piece of evidence would constitute a pattern of empirical facts that would lead them to overturn their preconceptions of intention and character. Of course, that might not always be possible. If I am convinced that evil entities are ready to strike at any moment, someone pointing out that no such entities have been observed will not sway me. On the contrary – their tendency to stay just beyond sensor range is further proof of their ill intentions…
But it is critical that we try to do so, as in a modern media environment, we are constantly evaluating scenarios we know nothing about, but which we then use to reinforce our preconceived ideological notions. A recent example of this is the horrendous Ebola outbreak in western Africa. After it became clear that American scientists had developed an experimental serum that was a potentially effective treatment, they were immediately accused of withholding it from the Africans, out of sheer racism. Once they considered making it available, they were accused of using this epidemic as an opportunity to test their half-baked drugs by experimenting on the poor Africans, again out of sheer racism. It is worth stating explicitly: There is nothing people who actually work on improving the human condition by developing cures for terrible scourges of mankind can do to dispel the notion that they are actually terrible racists, once these accusations are raised.
This situation is most acute on platforms like Twitter where people from the entire world communicate with each other, but often with little prior exposure and with an extremely limited shared basis of empirical facts. It can be hard to tell what someone is trying to say in 140 characters. Germans would call it the “Aussageabsicht”. Consequently, we observe plenty of misunderstandings and vitriol. With such a thin veneer of data, one can ascribe virtually any intention or character to anyone, regardless of what they say. And people do.
This is a dangerous situation and more likely to promote the bad than the common good in the long run, unless we reign in the tendencies brought about by our social neuroscience circuitry that is operating far outside of its design range (when it comes to social media). It would be a good start to admit that one does not really know the other people making the statements we object to and that judging them on this basis is really not a good basis for judging them as a person on the whole. On the contrary, we should be mindful that if we are willing to do so on a whim is both unwise and reflects poorly on us as an arbiter of empirical facts. Second, it is critical to recognize that any perception of intention or character is an inference and an inference that is made – given that we know next to nothing about the person – mostly on the basis of our prior beliefs, based on social norms and expectations and on ideology. It might feel compelling, but it is not necessarily so. Third, we should ask ourselves what behavior of the person could possibly be inconsistent with our preconceived notions, then go about eliciting such behavior. Everything else could be considered unfair and willfully unjustified. Finally, we should also recognize that seeing someone else making an accusation of ill-intentions does not constitute evidence of ill intentions. Sometimes, they just constitute evidence of bad inferences on part of the person who wrote the tweet. Not all tweets are created equal. Some are written by people desperately trying to advance the human condition as best as they can whereas others are written by people without achievement or judgment. They should not be weighed the same. Of course, it can be challenging to tell which is which, on Twitter.
To realize that not all tweets are created equal is particularly important for “shaming tweets”, tweets intended to shame the target (how one can ascertain that shaming is indeed the intention is the topic of an elaborate meta-discussion). It is important to recognize how devastatingly effective shaming is. Most (not all!) people are wired to protect their reputation at (almost) all cost. This is critical for autonomous agents that evolved to live in a social group. All social groups are beset with persistent principal agent problems. Reputation is one way to make sure that everyone does what they are supposed to. This is devastatingly effective. Whereas the concept is a little murkier in a modern global society, there can be no doubt about the fact that people’s behavior still changes dramatically depending on whether they think they are anonymous or not. But historical examples are most striking While most countries in WW1 simply drafted soldiers for WW1, the United Kingdom held on to their volunteer system for as long as they could. This is a problem if the country is in dire need for manpower, but few people are willing to sign up to fill this need. The example of the “Order of the white feather” shows how effective shaming can be. Men can quite literally be shamed to “volunteer” to fight and – not that unlikely – die or get maimed in the process. Yet, many picked this fate over being shamed. The pillory also comes to mind. In an age where hanging was the norm in terms of punishment for misbehaving and carefully calibrated torture was routinely used in criminal investigations, there was also an entire catalog of punishments specifically designed to shame people, the so-called “Ehrenstrafen“. It is not a coincidence that we abolished those along with torture – and for the most part – capital punishment.
One should recognize how devastatingly effective shaming actually is. It is so effective that there are entire movements with the sole position of emphasizing that “fat shaming” or “victim shaming” is not ok. In addition, Twitter allows for a kind of asymmetric warfare. Anonymous or pseudonymous twitter users can attempt to to shame whoever they feel like, for any reason and for the entire world to see, preserved forever. That is a tremendous power with no checks and balances whatsoever. Anyone trying to shame someone on Twitter for perceived misdeeds is effectively judge, jury and executioner all in one. Where are these checks and balances supposed to come from? The only way I can see this happening is from an evolution of internet culture itself. Perhaps people will learn – as they get more comfortable with the medium – to recognize that not every allegation or insinuation is equally justified and that some are more telling about the mental state of the shamer than reflective of the presumed misdeeds of the shamed. Not everyone with an internet connection and a twitter account has something insightful to say. People like that can’t be stopped, but can they be ignored?
Meanwhile, can we all try to stick to the known facts? We might all be surprised how little we can actually ascertain for sure. But it might lead to a healthier and less vitriolic online culture.