A painted smile decides nothing: an emotion has to be computed

For a long time, an emotion only had to be recognised. You looked at a face and matched it to a word. Joy, anger, fear, disgust: a dozen names, each with its own typical grimace. That was enough for as long as the task was to identify what someone else was showing.

Then the task changed. A system needed to have an emotion of its own and make decisions on the basis of it. At that point it turned out that the mask on the face decides nothing. Names were no help. The system needed numbers.

This shift sounds like a story from a robotics lab, and in part it is. But it describes a mistake that marketers, sales teams and managers make every day: they read the expression and assume they have read the state. For a company selling into Russia from abroad, that mistake is more expensive than at home, for reasons I will come to.

Stage one: names and faces

Historically, the first approach to emotions was linguistic. Researchers looked for names and mapped faces. The question was framed like this: here is an expression, which word fits it? That is where the classic work on basic emotions comes from, a small set of universal facial expressions that people recognise in much the same way across cultures.

The approach works, within limits. It is good for cataloguing, for telling one expression apart from another. What it cannot do is say anything about what is happening inside the system that produces the expression. It gives you no way to predict what that system will do next.

Most of the commercial tools a Western team meets today still sit on this first stage. Facial coding in ad testing, emotion tags in call analytics, sentiment labels in social listening: they take an outer signal and attach a word to it. The word is useful for a report. It is much less useful for a decision.

Stage two: attempts to calculate

The next step brought formulas that tried to derive the intensity of an emotion from measurable components. One example is the ratio between a need and the probability of satisfying it. Readers from the Russian tradition will recognise the line of thought associated with the Soviet physiologist Pavel Simonov, who described emotion as a function of need and of the information available for meeting it.

This is already a bridge from a dictionary to a calculation. The emotion is no longer only named; it is derived from other quantities. Classifications that linked a spectrum of emotions to traits of temperament through a wheel of categories played a similar transitional role. They tried to connect the dictionary approach to a structural one.

Still, the formulas of this stage described emotion as the result of a calculation while the calculation itself rested on a finite list of names. They did not yet describe the state of a system on continuous scales.

Stage three: a system that has to decide

The real shift happened in computational neuroscience, and a practical task pushed it. Engineers needed a robot that would adapt to a changing environment by itself, rather than display emotions on a screen for an observer. A screen with a painted smile decides nothing. It entertains the person watching. It does not help the system choose its next action.

To adapt, the system has to assess its own state in terms of scales:

• how large the gap is between where it stands and its goal;
• how mobilised it is;
• which way the flow points, inward or outward;
• how much resource remains.

These parameters are continuous, unlike the discrete names in a dictionary, and the next action can be derived from them directly. The word “fear” tells a robot nothing. The parameters “high intensity, vector inward, forecast of no result, full immobilisation” tell it exactly what to do: freeze, conserve resource, wait.

Why the mask is not the emotion

The expression on a face is not an emotion. An emotion is a state of a system, described by the parameters of a process, and the outward expression only accompanies it. The two can fail to match entirely. People smile through inner anxiety. They keep a level face during an inner storm. If you rely only on what is visible, a wrong reading of the state is guaranteed sooner or later.

This explains why the dictionary and facial approaches fall short wherever you need to understand what is really going on inside, whether inside a person, a team or a system. A name describes the form from outside. A scale describes the mechanism from inside.

Two identical faces at one meeting

Take two people at a meeting, both with the same neutral expression. The facial approach gives one conclusion: both are calm.

The scale approach can give two very different pictures. The first person really does have a low-intensity process and an open, outward vector. They are relaxed. The second has a high-intensity process with the outward expression fully suppressed. Their system is at its limit, and none of it shows.

Judged by the face, both land in the same category. Judged by the scales, the difference becomes visible long before anything breaks through to the surface.

Where Russia makes this sharper

Here is the part that matters for a foreign company.

In many Western business cultures, and especially in American ones, the face carries a lot of social signalling. A smile at the start of a call, visible enthusiasm in a demo, a friendly expression from a buyer: teams learn to read these as data. Much of Western sales training is built on that reading.

In Russia the default is different. A neutral face in a business setting is common and carries no negative meaning. A smile without a reason can read as insincere rather than warm. So a foreign sales team that reads Russian counterparts with the habits it formed at home will make two errors at once. It will see coldness where there may be real interest, and it will see calm where there may be tension it ought to address. The neutral face from the meeting above is not an edge case here. It is the ordinary picture.

The second difference is the channel. A large share of business contact in Russia moves into messengers, so a growing part of the conversation with a buyer happens with no face at all. A team that depends on expressions loses its main instrument. A team that knows how to read intensity, direction and resource can read them in text too: in how fast someone replies, in whether their questions point at your offer or at their own internal problems, in whether they are still asking or have started to withdraw.

The third difference is how research answers behave. Survey labels such as “satisfied” or “interested” are already a dictionary answer, and a respondent who keeps a neutral face in person may keep a neutral answer on a form as well. A label collected that way describes the surface. It says little about whether the person has the resource to act or is already mobilised against you.

None of this requires a theory of national character. It requires noticing that the instrument you trusted at home was calibrated on a different surface.

What this changes in practice

If you want to understand the real state of a buyer, a team or yourself, asking “how do you feel?” and getting a name is not enough. It is more useful to go through the parameters:

1. Intensity. How strong is the process right now? Is something at stake, or is this a routine choice?
2. Direction. Where does the flow point? Outward, towards your offer and the conversation, or inward, towards internal problems, approvals and risks?
3. Resource. Is there anything left to spend: time, budget, attention, political capital inside the company?
4. Mobilisation. Is the person gathering themselves to act, or have they gone still?

A name is a label on a box. The scales are what is inside it.

For a company entering the Russian market, I would turn this into three working habits.

Brief your local team in parameters, not labels. “The client seemed positive” is a dictionary report. “The client asked three questions about implementation timing and none about price, replies within the hour, and has started bringing in a colleague” is a report on intensity, direction and mobilisation. The second can be acted on.

Rebuild your research questions. When we run customer research for foreign companies, the interview guide avoids questions that invite a label. It asks what was at stake, how urgent the situation was, where attention went and what the person had left to spend on the decision. Those answers map onto the scales.

Do not import your reading of faces with your sales playbook. If you are preparing a market entry into Russia, treat expressions in meetings as weak evidence and behaviour over time as strong evidence. The quiet buyer is not necessarily cold. The polite buyer is not necessarily ready.

Scales before vocabulary

This discipline is built around scales rather than around a vocabulary of emotions, however familiar the vocabulary is. The vocabulary is good for talking about an experience after it has happened. Scales are needed to understand what is happening while it is still happening, and to do something about it while that is still possible.

A robot that only paints a smile on its screen never learns to act. A sales team that only reads smiles never learns what its buyers are going through. In both cases the fix is the same: stop asking which word fits the face, and start asking what state the system is in.

Frequently asked questions

Sources

• Russian version of this article on belokrylovo.ru: Роботу нельзя нарисовать улыбку: её нужно вычислить