Artificial intelligence with a human face. How neural networks build emotional connections with customers

Second, emotional attachment fosters repeat purchases. People love to return to those brands which previously brought to them some nice feelings in dealing with them. This is especially important in developed markets, where many products and services are of almost the same quality. Under these circumstances, reasoning gives way to personal emotions. Consumers continue purchasing from the company that brings some kind of satisfaction to them despite the presence of some alternatives having the same specifications.

Social aspect also kicks in here. When a brand is capable of evoking strong emotions among its consumer group, it goes beyond its position of a product and enters the sphere of public discourse. People are more likely to comment on such brands, discuss them on social media, and recommend them to others. What follows is a phenomenon of free marketing, where information about the product is shared not through advertising campaigns but through people voluntarily advocating for the brand.

Thus, emotions become the secret to establishing long-term brand loyalty. They establish a unique user experience that not only helps the brand remain in the memory but also makes it an enriching aspect of the consumer’s life. While in a very competitive market, brands that are capable of making their viewers feel true emotional connections gain a significant advantage because loyalty induced through emotions is much more robust than commitment made based on mere logic.

1.2 The role of personalization: from CRM systems to neural networks

Personalization in marketing, as an approach to adapting products and services to the unique needs and preferences of individual clients, began to develop in response to the demand for deeper and more effective consumer interactions. Its evolution has gone through several stages, starting from the early use of databases to modern high-tech solutions based on AI.

The first serious attempts to personalize marketing and customer service can be traced back to the introduction of Customer relationship management (CRM) systems in the 1980. At that time, companies began to recognize the importance of effective customer engagement and the need to store data on each client’s needs, preferences, and purchase history [20]. Such systems became the first significant step in this process. They allowed companies to track data on transactions, customer preferences, and brand interactions, enabling the creation of targeted offers and the optimization of marketing efforts (fig. 7).

Figure 7. Personalization process using CRM systems

However, despite their importance, CRM systems were limited in terms of flexibility and accuracy. Typically, simple market segmentation methods were used based on basic customer characteristics, such as purchase frequency or product types. Within this approach, personalization remained at the level of basic offers tailored to user groups rather than to each individual client. Their main function was to collect data, not to conduct in-depth analysis. This was an important step toward personalization, but it still did not provide sufficient depth in processing information to predict future customer needs.

In the next stage, personalization took a new form through automated marketing. This became possible thanks to the development of platforms capable not only of storing and processing customer data but also of initiating communications with users based on their actions. During this period, digital marketing began to shift away from universal mass mailings toward more flexible and targeted interactions.

The newly introduced tools enabled automatic messages to be triggered in response to specific events in the user environment. If a customer made a purchase, browsed certain categories, or abandoned a shopping cart, the system could independently respond with an appropriate message – whether a thank-you email, a reminder, or a personalized offer. This approach significantly increased the relevance of communications, as the messages were based on actual consumer behavior rather than assumptions.

However, despite the obvious progress, these scenarios remained predictable and static. They relied on pre-defined templates and could not adapt to changes in context or customer behavior in real time. Interaction with the user was based on simple cause-and-effect relationships, without accounting for nuances such as motivation, temporal preferences, or emotional state. The logic of personalization within automated marketing was limited by pre-established conditions and did not include the system’s ability to learn or self-update.

Nevertheless, this stage marked an important milestone in the history of personalization. It allowed marketing systems to move beyond data storage and toward the active use of data in communication processes. Moreover, it was during this period that the concept of sequence and continuity in the customer journey was introduced – where each client action could prompt a response from the brand. This laid the groundwork for further algorithmic complexity and led to the transition from template-based solutions to intelligent personalization systems capable of analyzing behavior in a broader and more contextually rich framework.

As development continued, the technology reached a qualitatively new level due to the increasing use of machine learning algorithms in marketing strategies. Unlike previous stages, where user interaction was based on fixed triggers and linear scenarios, the new approach allowed for the analysis of complex dependencies between various behavioral parameters and the delivery of individualized solutions in real time.

Recommendation systems became the core of this transformation. Their operation was based on the ability of algorithms to detect patterns not only in the behavior of a single user but also across massive datasets accumulated from millions of others (fig. 8).

Figure 8. The principle of how recommendation systems work

This made it possible to build personalization not only based on direct indicators – such as «what you purchased» – but also through similarity with other users who performed similar actions, searched for similar content, or displayed comparable activity on platforms. As a result, content, offers, and communications began to be shaped by probabilistic models that take into account the statistical closeness between behavioral patterns.

This type of personalization proved especially effective in areas where the choice of a product or service involves a high level of uncertainty – for example, in entertainment, e-commerce, or streaming content industries. Platforms could suggest movies, music, or products even before the user consciously recognized the need to make a choice, thereby shortening the decision-making process. Moreover, the system considered not only past behavior but also the context of the current session – day of the week, time of day, device used, and duration of interaction. All of this enabled a more finely tuned communication, where the user received suggestions tailored not just to their general interests, but to their current state and expected actions.

User interaction became continuous and synchronized across different touchpoints: website, mobile app, social media, and email. Behavior in one channel influenced recommendations in another, creating the sense of a cohesive, individually tailored customer journey. Personalization ceased to be fragmented and evolved into a systemic adaptation of the brand to the unique rhythm of each customer’s life.

This became possible not only due to the growth of computing power and the accumulation of large volumes of data but also thanks to the increased accessibility of cloud storage and data processing technologies. Companies gained the ability to collect, synchronize, and analyze user data from multiple sources – something that previously required expensive infrastructure solutions. As a result, personalization evolved from a tool for local interaction into a full-fledged mechanism for strategic customer experience management, where every step was planned, predicted, and supported by data.

In the next stage, personalization continued to evolve, shifting from analytical behavior prediction to deep adaptation and interpretation of user context. A key development was the large-scale transition to systems based on neural networks and deep AI architectures. This made it possible not only to process vast datasets but also to work with their internal structure at a new level of complexity, uncovering subtle patterns in behavior, motivation, and emotional responses.

At this stage, personalization became multimodal – it combined textual, visual, auditory, and behavioral data to build a multilayered, dynamically updated user profile. For example, the analysis of text queries in search engines, combined with image or video viewing, as well as data on content reactions (viewing speed, scrolling depth, clicks, pauses), began to be used collectively to more accurately predict interests and intentions. This is no longer just a reaction to behavior – it is proactive modeling of intentions, in which the system strives not only to understand what the user wants, but also why they want it, what state they are in, and which forms of communication will be most appropriate.

Natural language processing (NLP) technologies gained significant importance at this stage. Systems learned not only to recognize queries but also to understand context, intonation, and even the emotional tone of a message. This enabled the creation of personalized dialogues, where the communication style was adapted to the user based on their speech patterns, writing style, or voice responses.

A major shift also occurred as personalization began to take into account not only behavioral and demographic characteristics but also more nuanced psychographic and emotional data. Marketing practices started to incorporate mood and emotional state analysis technologies based on the evaluation of text, audio, video, or even facial expressions. This allowed systems to adapt not only the content of offers but also their presentation: tone, visual design, timing, and delivery channels.

During this period, personalization acquired true cognitive capabilities – that is, the ability to learn, interpret, and predict. Every interaction with the user was used to refine the model of their preferences, interests, and potential needs. This allowed brands to build evolutionary models of interaction that developed in parallel with changes in user behavior, rather than remaining fixed at the point of initial contact.

Interface flexibility played a special role at this stage. Personalization systems began to influence not only content but also form: visual elements, page structure, and layout blocks changed depending on how the user perceived information, what device they were using, their mood, or the specific tasks they were solving at that moment.

Thus, before entering the modern stage, personalization evolved from behavior prediction to contextual and empathetic adaptation based on a deep understanding of the individual (table 2) [21].

Table 2. The evolution of personalization

Thus, this is not merely technological automation – it is the transformation of personalization into a form of cognitive interaction between the system and the user, in which decisions are made not only based on logic but also by considering psychological, situational, and even ethical factors. However, these advancements have also brought new challenges.

Modern personalization requires significant computing power and large volumes of data, making it increasingly complex and multi-layered. It is also essential to consider the need for data privacy. In an environment of constant changes in consumer behavior and intense market competition, personalization becomes not only a means of attracting customers but also a necessary tool for retaining them. It enables brands not only to respond to requests but also to create unique and valuable interactions that significantly influence brand perception and market position.

With technological progress, market globalization, and the rapid expansion of digital platforms, customer needs and expectations have undergone substantial changes. In a world where nearly every aspect of daily life is becoming increasingly digital, consumers are no longer satisfied with standard products or services. Instead, they increasingly seek personalized experiences that precisely reflect their individual needs, interests, and lifestyle.

The shift from mass service to an individualized approach has become one of the most significant changes in consumer behavior models over the past decades. This shift is driven not only by the development of technological capabilities but also by changes in how the value of interaction between brand and customer is perceived. There are several reasons underlying the formation of such expectations (fig. 9).

Figure 9. Reasons behind the formation of personalization

The primary reason for this is the rapid development of technologies that enable brands to collect and analyze customer data, as well as the evolving dynamics of interaction between consumers and products. Brands began offering suggestions not only based on past consumer behavior but also on their expectations and current mood. This approach was immediately perceived by users as convenient, allowing them to avoid unnecessary effort in searching for the right content or product. Gradually, this expectation became the norm, extending not only to media platforms but also to various business sectors – from banking services to online retail.

Another factor that shaped this expectation was the information overload faced by modern consumers. In a world of pervasive digitalization and saturated information spaces, consumers are now confronted with an overwhelming number of products and services. In such a context, only the brand that can address individual needs is able to capture their attention.

Additionally, since technologies developed and customized solutions came within reach, these practices had come to be regarded as an obvious thing. Sites like Google and VK that gather enormous amounts of user activity data have been able to get ahead of the users’ wishes and provide just the content that they will want most. This not only made it easier for these services to be used but also affected people’s expectations from all the brands they interact with. People began expecting that businesses ought to provide them with not just homogenous solutions but also unique solutions that reflect their personal taste.

Yet another important contributory factor to the demand for a personized approach is having the sense of being special and appreciated. Where consumers are often overwhelmed in a mass production culture with the same goods and services, they desire to experience that their specific needs and interests are applicable. Personalization can capitalize on this sense of being appreciated, where it can emphasize to consumers that the offer and experience is based on their own, exclusive data. This approach impacts not just loyalty but also emotional brand perception. When customers feel appreciated and listened to, it creates a firm connection with the brand – one that goes beyond consumer interest and is a deeper psychological attachment.

In addition, over the last several years, with growing collective consumer consciousness regarding personal data and privacy, consumers now expect not only that their data will be used by brands to personalize, but that it is done so in a manner respectful of their right to privacy. With personalization becoming the norm, customers find that they are demanding to be provided with some level of insight into how they are handling their data and that they would not trade it to third parties without their consent. This compels customers to push brands not only to apply personalization technologies but to ensure that they are as safe and ethical as possible to users.

The call for personalization, then, is not merely due to a desire for speed and convenience but also because customers increasingly seek not merely a product or service but an experience that resonates with their value system, interests, and lifestyle. The trend calls on companies not merely to adopt new data-collecting and data-analytical technologies but also to be capable of evolving in reacting to shifting demand and taste. In a competitive business environment, those that fail to meet expectations of personalization are likely to fall behind global trends and lose the competitive edge.

Conclusion to chapter 1

Emotions are a key part of marketing and play a significant role in consumer behavior and decision. Neuropsychological studies prove that emotions not only govern short-term reaction but also long-term brand commitment, dramatically increasing customer loyalty. Emotional response is a much more powerful driver of decision than reason or fact, as a variety of research and extremely successful ad campaigns such as those of Nike and Coca-Cola have clearly illustrated. Feelings form connections that remain in the memory for quite a long period and influence the choices of consumers even if they are presented with products of similar kind.

This not only increases sales but also helps build a solid rapport with the brand, producing a sense of engagement and uniqueness in customers. Through the use of emotions in marketing, companies can distinguish themselves from others by positioning their brands as something desirable and consumer centric.

With technological advancement – more particularly the development from basic CRM systems to neural networks – marketing has been made increasingly personalized. It has become a key tool not only for customer acquisition but also for customer retention in a highly competitive marketplace. This development has, however, brought with it challenges related to handling extremely large data and ensuring customer privacy.

As a result, the need for an individualized approach has become a normal component of consumer behavior. Consumers now anticipate not just products and services, but an experience that is suited to their lifestyle and interests. Those companies that can deliver such an approach are already a step forward in a world of global change and rapidly changing market demands.

Chapter 2. AI and emotions: how it works

With the development of AI, one of the most ambitious tasks has become the understanding and analysis of human emotions. Unlike the traditional approach, which often boils down to superficial data processing, modern technologies allow for a deeper comprehension of feelings. As a result, it has become possible to create higher-quality, personalized user interfaces and improve the customer experience. Emotions play no less an important role than information in the process of decision-making, interaction with a product or brand, and therefore the ability of technical devices to perceive and interpret them becomes especially important.

State analysis technologies based on AI are capable of recognizing even those emotions that are hidden behind the external manifestations of human behavior. They are actively used in various fields: from creating personalized recommendations on streaming platforms to improving customer interaction. Their importance is hard to overestimate, as they provide more natural and human-like ways of interaction.

2.1 Neural networks as emotion readers

Modern research in the field of AI shows that emotions have become not only an object of study but also an important element of interaction between humans and machines. A significant question remains how exactly it is capable of «reading» them through the face, voice, and text.

One of the most studied and applied methods is emotion recognition based on facial expression analysis, allowing a person to identify the state of an individual with accuracy. Facial expressions are one of the most expressive and direct indicators of internal emotional states because they have a tendency to show emotions unconsciously and involuntarily. This makes them highly valuable in real-time interaction where voice or verbal cues are unavailable or unreliable.

Scientifically, facial expressions are based on universal muscle movements, as can be seen in the research of Paul Ekman and Wallace Friesen, who developed a system of facial action encoding. It defines specific units of action that build up into emotional expressions that are perceivable. For example, a genuine smile requires the contraction of both the big zygomatic muscle (pulls mouth corners upwards) and the orbicular eye muscle (creates «crow’s feet» around the eyes). It is difficult to fake these physiological reactions consciously, and therefore, they are very informative for emotion recognition systems.

The working principle relies on the observation of various facial features, such as the shape of the eyes, the angle of the lips, the movement of the eyebrows, and other small changes that may indicate a particular state. These changes can be both visible and not detectable by the normal human eye. For their effective recognition, various algorithms are used, each with its own features and applicability in different contexts.

One of the most widespread tools is the convolutional neural network (CNN) [22]. These neural network algorithms, based on the principles of machine learning, enable deep image analysis by extracting various features that play an important role in recognizing facial expressions (fig. 10).

Figure 10. Architecture CNN

They efficiently process visual data, automatically extracting and classifying features such as contours, textures, eye shapes, and corners of the mouth. At a low level, they detect simple elements, while at a higher level – complex patterns such as the shape of facial expressions, which allows for accurate classification of emotions such as joy, sadness, anger, and others. In addition, CNN are spatially invariant, i.e., they can identify expressions of emotion appropriately even when the face is rotated, partially occluded, or slightly shifting in location. They are therefore appropriate tools for real-world applications such as emotion tracking, adaptive user interfaces, and human-computer interaction systems. To improve recognition accuracy, recurrent neural networks (RNN) and long short-term memory networks (LSTM) are often used, as they enable modeling of temporal dependencies (fig. 11).

Figure 11. Architecture RNN and LSTM

They can analyze the dynamics of facial expressions by tracking their changes over time, which is especially important in real-life situations where emotions can shift during interaction. For example, recognizing them based on video clips requires accounting for temporal aspects, such as changes in facial expressions during a conversation. They are built to take context and event order into account, and thus are particularly well-suited to dynamic data. In addition, multimodal solutions guarantee a much better accuracy of recognition. Emotions are rarely depicted by the face alone, most of the time they are complemented by voice, gestures, body movements, and physiological states. These models integrate, for example, visual data with auditory data processed through speech parameters and create a richer picture of the emotional state. This procedure helps in preventing interpretive errors and accommodating better the user’s individual and cultural features.