Editorial note: This series presents Psychological Engineering, an original framework developed by this site’s authors. It is a perspective piece exploring ideas at the intersection of psychology, systems thinking, and everyday life — not an established psychological theory or academic standard.

Chapter 3: Directional Emotions

Humans are social animals, which means that evolutionary selection does not take the maximization of the individual's probability of survival and reproduction as its sole goal. Human individuals have evolved attitudes toward different types of individuals: fear of a powerful enemy (corresponding to the flee decision), hostility toward a weak enemy (corresponding to the attack decision), the attitude toward offspring, the attitude toward a mate, the attitude toward elders.

These "attitudes" mean that human decisions manifest social characteristics, and are also closely related to the continuation of genes. We can imagine two tribes:

In one tribe, everyone lives for themselves
In one tribe, there is organization and discipline, and people help each other

The latter's probability of continuation in a harsh environment will far exceed that of the former.

This is the evolutionary significance of directional emotions. Emotions such as liking, aversion, reverence, fear, and love, on the surface, are feelings directed at a certain type of object, but behind them correspond to behavioral tendencies directed at this type of individual.

Next, we enumerate the most important directional emotions and their influence on corresponding decisions:

1. Reverence & Submission

The directional emotion of "reverence" creates imperative utility, which is part of the composition of utility and controls the formation of a submissive or independent personality.

We can understand that evolutionary selection would preserve such a property that appears at random: the individual can recognize in the group a person who is confident, considerate, has a past record of successful decisions, and considers the group, and forms a reverence directional emotion toward them.

In primitive tribes, such people mainly fell into two types:

The elders of the family
The leaders of the group

Individuals in the group will form a directional behavioral tendency toward such objects because of the above characteristics, and the feeling of the reverence directional emotion is the feeling attached to this directional behavioral tendency. This is the source of imperative utility.

How much an individual's imperative expression toward another individual (that is, an order) can influence the other's decision, that is, the strength of the imperative utility, is related to the degree of "reverence."

Therefore, our directional emotion — the feeling of awe and reverence — is an emotional feeling, but behind it corresponds to the influence of the corresponding emotion variable on our decision, that is, the imperative utility. A parameter controls the degree to which the imperative utility takes effect, and thus controls a submissive or independent personality.

2. Friendliness & Altruism

On many higher animals we see the characteristic of "altruism," the typical case being a mammalian mother who has just given birth to her child: the arduous process of breastfeeding, and giving up fleeing in the face of a powerful predator to protect the cub... these are portraits of "altruism."

Directing altruism at one's own child is directly related to whether genes can continue, so altruism directed at offspring is preserved by evolution. Secondly, "altruism" toward other family members or other individuals in the group makes the family and the group unite to face external threats, increasing the probability of survival of each individual in the family and the group.

Let us see how to realize altruism engineering-wise:

After evaluating the utility of the event's counterpart, according to the directional emotion toward this object of "love" or "friendliness" (we argue that "love" and "friendliness" are different words corresponding to different degrees of the same directional emotion), a part of the utility is incorporated into one's own total utility evaluation. In this way, the other's interest is incorporated into the evaluation of one's own decision, creating altruism.

When the conversion ratio of utility exceeds 1, the individual's behavior will show the characteristic of self-sacrifice. For example, the conversion ratio for the utility of offspring is very high, so one will make a choice whose negative utility to oneself even exceeds the positive utility to the offspring.

We can control the ratio of the other's utility converting to the AI's own utility through a parameter:

High ratio: the altruistic plot in the personality is heavier, more inclined to help friends, and make self-sacrifices for family
Low ratio: the altruistic characteristic in the personality is faint, and may even behave selfishly toward one's own relatives

Topic: Self-Serving and Pure Altruism

There is a kind of "altruism" appearance that is easily confused with true "altruism," and their formation mechanisms are also different.

For example, some people's altruistic behavior is aimed at exchange as its purpose. When we educate children that "those who help others are more welcomed by everyone" and "if you are good to others, others will be good to you," the essence of this information is to create causal knowledge between behavior that has the appearance of altruism and other behaviors (events).

In this way, the goal-shift function of the cognitive system will take effect, and the emotion system will shift the event goal of "being welcomed by everyone" and "hoping others are good to me" onto the event class of "helping others."

For many people, the idea of helping others is deep-rooted, yet they forget what the initial reason was. They enjoy the benefits of altruism, and mistakenly believe themselves to be selfless people. But the root of these "altruistic" behaviors is the self-serving original motive.

And true "altruism" is, as said above, incorporating the other's utility into the evaluation of one's own decision utility, very pure, and its corresponding directional emotion is "love."

3. Hostility & Harming Others

If "altruism" is the emotional reaction of the directional emotion of "love" and "goodwill," then correspondingly there is "harming others," which is the emotional reaction of the directional emotion of "hatred" and "hostility."

On the primitive continent, the evolutionary significance of "harming others" lies in that if hostility can reasonably be directed at the individual who forms a threat to one's own survival and reproduction, then events harmful to this individual are likely beneficial to oneself.

For example, males form great mutual hostility when competing for mates, to the point of fighting at the cost of their own lives. We can imagine that if the individual did not have this emotional reaction, it would certainly be eliminated — because every time it encountered competition, approaching benefit and avoiding harm would certainly make it choose to yield and avoid fatal fighting, and the chance for genes to continue would be lost.

The engineering realization of "harming others" is exactly the opposite of "altruism":

After evaluating the utility of the event for the counterpart
According to the directional emotion toward this object of "hatred" or "hostility" (we argue that "hatred" and "hostility" are different words corresponding to different degrees of the same directional emotion)
Incorporate a part of the utility into one's own total utility evaluation, and:
- The counterpart's positive utility serves as one's own negative utility
- The counterpart's negative utility serves as one's own positive utility

In this way, the other's interest is incorporated into the evaluation of one's own decision, creating "harming others."

When hostility deepens to the point of hatred, the individual's utility conversion ratio exceeds 1, and at this time the individual is even willing to make great sacrifices to exchange for harm to the counterpart.

And attack, including behavioral attack and verbal attack, is the most common example of the "harming others" reaction caused by hostility: the individual is willing to bear conflict and other risks to create harm to the counterpart.

Like altruism, we can control the ratio of the other's utility converting to the AI's own utility through a parameter:

High ratio: the AI is more prone to hostility, more prone to aggression and revenge
Low ratio: the AI's aggression is weak, not prone to hostility, and prone to tolerance

Topic: Envy and Jealousy

When we recognize a positive event of the counterpart:

We put ourselves in the subject's place and evaluate the utility for ourselves
If this imagined event has high utility, we will think "I really wish this event would happen to me too"
This is the source of envy

And jealousy is a compound feeling:

On the one hand, one wishes this event would happen to oneself
On the other hand, one forms a negative emotion upon hearing of a rival's positive event (equivalent to one's own negative event)

4. Sympathy

When an individual recognizes that the counterpart is in a situation of extreme negative emotion, because of the counterpart's negative emotion a motive to help is generated, and the corresponding emotional feeling is sympathy.

Motive and feeling correspond, so humans will regard the seen behavior of helping a pitiful person as the individual having a sympathetic disposition.

Somewhat similar to altruism, the difference lies in:

They arise from different directional emotions
The altruism reaction targets the directional emotion of "love," so only toward offspring, relatives, lovers, and close friends is there such altruism
Sympathy is not limited by the directional emotion, but humans do indeed more easily feel sympathy for certain types of objects, such as weak and cute objects, and children

Engineering-wise, we can control through a parameter the degree to which the counterpart's negative situation is turned into sympathy and a helping motive. Controlling this parameter:

Can create a sympathetic AI: she not only feels sympathy and the impulse to help toward people in miserable situations, but is also more reactive to complaints, and will hesitate, soften, and easily forgive because of the negative effect of a decision on others
Can also create an indifferent AI: no feeling toward others' miserable situations, listening only to the information of complaints with no emotional reaction, making decisions without considering negative effects on others (unless they re-affect oneself), not prone to softening and forgiving

So far we can see: the empathy ability is at the cognitive level, and an individual can have empathy ability but no sympathy.

The directional emotion of pity creates the sympathy reaction, and correspondingly the directional emotion of aversion creates the bullying reaction. It has the same formation mechanism as the sympathy reaction, only the utility of bullying is converted to negative, that is, the individual bears negative utility to create harm to the aversive object.

Likewise a parameter controls the utility conversion ratio of the aversion directional emotion:

When this parameter is high, the individual is more prone to "bullying" the aversive object
When this parameter is low, the individual's bullying tendency is weak

5. Fairness Utility

Fairness utility balances crime and punishment, compensation and reward. For the individuals in a group:

What degree of punishment the individual who did wrong should suffer
What degree of compensation the individual who bore extra harm should receive
What degree of reward the individual who made a contribution should receive

These will form an inner kind of fairness recognition. Although each individual's specific conceptions differ, a relatively consistent group identification of "how to be fair" will eventually form. This is a very important balancing identification in a group, maintaining group order. After this group identification is routinized, it influences the formation of law.

The individual's sense of fairness, whether manifested in individual behavior or in group behavior, its effect is all a kind of choice utility, participating in the competition of the utility model. Because it is manifested as an attitude toward the individual, we discuss it in this chapter on directional emotions.

In terms of engineering realization, the individual needs to form a fairness conception at the cognitive level, for example:

According to the degree of the wrong done by the object, and the degree of harm formed to oneself, what degree of harm should be borne
The harm the object has already borne in the individual's cognition needs to be compared with this degree, to determine the motive of punishment and compensation

That is to say:

If the individual has only borne a very small harm, then the fairness utility of the motive to make the individual bear harm will be high
If the individual has borne a great harm, then the fairness utility of the motive to make the individual continue to bear harm will be low
If the individual has borne harm exceeding this value, then the motive to compensate the individual will gain fairness utility