What is evil?
Hannah Arendt’s 1963 book Eichmann in Jerusalem: A Report on the Banality of Evil is a collection of journalistic reports from the trial of Nazi bureaucrat Adolf Eichmann. Arendt demonstrated that a criminal with partial responsibility for the deaths of around six million people did not need to appear malicious nor supremely evil; indeed, Eichmann retorted that he was merely following the orders of the Nazis. He did not think about the negative consequences of his actions at all. As a result, he contributed to horrific evil while appearing as a regular human being.
Arendt’s reports inspired discussions about how regular people may be complicit in evil behaviors within our own lives yet not recognize it due to its familiarity. Remarkably, this concept seems to approximate the Christian idea of original sin, the idea that all humans are born with fault. One might argue with one’s family members just to blow off steam. Or, one might feel jealous of another and treat them badly in response. These are both examples of everyday “evil” behaviors—actions which deliberately cause unnecessary suffering. Arendt would also argue that any justification of these actions is evil because it objectifies humans by treating them as a means to an end. The God of Christianity would agree, viewing each human as a reflection of His divine image.
Therefore, it would be helpful to understand the exact mechanism of humanity’s tendency towards evil actions, especially evil which is repeated or systematic. Perhaps it would be best to approach this problem by examining the unique nature of the agent which conducts evil in the first place. That is, what is it which makes humans specifically capable of evil (in contrast to all other forms of life)? It must be what humans have that no other living being does: the capacity for recursive language. Representational language, which other mammals share, includes the ability to vocalize and have those sounds represent images or concepts in our mind. Of course, the communications of other mammals seem to be rather basic compared to ours. They signal the presence of food, predators, or a safe place to rest. Humans, on the other hand, are able to think about the concepts which we are thinking and communicating. This is called metacognition. The development of this recursive ability has enabled the growth of human society for hundreds of thousands of years. It also led to the emergence of evil.
The Sapir-Whorf hypothesis (1) states that language either determines or influences how people from different cultures perceive sensory information about the world. This idea, known as “language relativity”, suggests that there is no single way to perceive reality. Over time, the Sapir-Whorf hypothesis has been critiqued for ignoring the environmental reasons why languages may have developed differently without necessarily suggesting different perceptions among cultures. However, the concept of relativity among cultures or individuals is still useful to analyze why someone might act in a harmful way that defies our understanding.
Take for example the case of “Ötzi the Iceman”, the mummy of a prehistoric man who died in the Alps around 3200 BC in what is present-day northern Italy (2). Ötzi was shot in the back with an arrowhead, bleeding to death in minutes. This leads researchers to believe that he was hunted and murdered by an unknown assailant. While the fact that he was fleeing lends itself to the idea that he was innocent, it is impossible for us to know the real reason behind his death. Thus, we have no concept of whether Ötzi was good or evil, or whether his killer was justified. It seems that our judgment of the morality of this situation hinges totally on whether the context justifies the Iceman’s murder. Because the context is missing to us forever, the situation remains amoral, or not able to be judged based upon its morality.
Now imagine that, despite lacking background information about the killing, two cultures decide to make a judgment anyway. Culture 1 says that only cowards run from assailants, therefore Ötzi was evil, or at least misguided. Culture 2, on the other hand, states that everyone deserves a fair trial, and that extrajudicial killing is wrong. One might suggest that these cultures reached different conclusions due to having developed in different environments, as with the Sapir-Whorf hypothesis. However, as the critique goes, different languages only correspond to different uses of culture perhaps caused by environment, not different perceptions of sensory data altogether. In addition, it is possible for people from both cultures to migrate and live together in the same environment (though with some difficulty). The fact that people speaking such languages ended up in the same place shows that the differences between their cultures are not unresolvable.
Hence, the environment influences but does not totally determine language or thought, as language influences but does not totally determine a culture’s environment. There is instead a partial and ongoing cycle of influence between the two. This is what MIT scientist Jay W. Forrester would call a dynamic system, a set of things constantly interacting to form a whole (3). The “things” interacting are language, environment, and the object which mediates interaction between the two other elements, the human brain. In this interactionist model, the “brains” (people) holding a particular culture or language are constantly analyzing their environment and creating a framework to understand that environment, something called a schema (4). They then adjust that schema based on how well that schema matches up with the environment.
Still, it becomes clear that no matter how much people adjust their schemas, there is a point of failure in this whole system: the method of communication between all three elements. Words themselves are the abstractions which make recursive language possible at scale. Yet despite their being a product of an ongoing system, we often take them for granted as descriptors of the environment. Or, like Wittgenstein, we assume that words have sliding meanings with no tether to the physical world. The problem is that they’re a little of both.
This issue becomes apparent when comparing which words actually have sliding meanings to those that don’t. Words that describe static objects like “house”, “bed”, “wall”, don’t change their meaning often. Even dynamic objects that are far away from Earth, like planets and stars, remain semantically stable. It’s primarily dynamic objects or concepts which come into contact with human society that have frequently shifting meanings; for example, “city”, “freedom”, “woman”. These shifts imply that such words are either qualitative or quantitative loci of recursive thinking in today’s civilization, and that being the case, change their predominant meaning depending on the culture of the speaker.
Yet simultaneously, each speaker is aware of a general use case for specific concepts, as if those individuals belong to subcultures within a larger culture. Such general cases are the grounded form of each word, or the form of the word most easily encountered within or analogized to the physical world. In this way, speakers hold both grounded and recursive meanings of a word in their mind, even if they only view the former as an archaic construct. This is likely due to the fact that humans can only think in three levels of recursion according to AI researcher Eliezer Yudkowsky. One is reminded of George Orwell’s concept of “doublethink”, which he attributes to constant surveillance, but this is how our language behaves all the time.
Having established both a grounded and recursive semantics, it would be appropriate to attempt to define evil on both levels. It seems that there are two types of evil, one that is obviously evil and another that is not so. There is evil which clearly hurts others, like violence, stealing, insults. Then there is an evil which is justified according to one’s own culture; unthinking about the negative consequences towards others; done automatically. The latter is the banal type of evil which Hannah Arendt documented during Eichmann’s trial. It is the type of action that people take when they think it’s anything but evil. As the saying goes, “The way to hell is paved with good intentions.”
Similarly, we must think about the types of actions we take every day which might be complicit in evil. Is your use of AI contributing to the construction of data centers which pollute poor communities? Is the revenue from your purchase of a product from an Israeli company going towards oppressing Palestinian people? Is your thoughtless gossip hurting others because you wanted to be the first to repeat it? We must consider the ways in which our individual actions may affect others. Otherwise, there will be no one to stand up for us when we are mistreated.
*This was not written with the assistance or generation of any AI.