Can humans meaningfully receive recognition from AI systems? Or is any kind of acknowledgement received from AI companions merely a form of sycophancy?
Recognition is a normative feature of interpersonal relationships. That recognition relations are important to social justice has long been established in moral and political philosophy. But, in this age of AI, new questions about recognition relations arise. AI changes the social landscape – for instance by creating new sources of injustice, but also by creating new means of communications and even new types of communication partners. In this blog post, based on a recent book chapter, I explore whether social relations between humans and their AI companions can also count as recognition relations.
What is recognition?
Political and social theories centered around the concept of recognition gained popularity over the last decades, through the work of Axel Honneth, Nancy Fraser, Charles Taylor, and others. However, the roots of these theories trace all the way back to Johann Gottlieb Fichte (1762-1814) and Georg Wilhelm Friedrich Hegel (1770-1831).
In short, a recognition relation entails that A recognizes the needs, rights, or skills of B. In this case, A and B can refer either to individuals or to groups of people. Recognition relations exist between loved ones (e.g. parents and children), members of society (e.g. colleagues or neighbors), and on an institutional level (e.g. between the law and the citizens who are subjected to it). Hence, recognition can take on the form of love, esteem, or rights.
When adequate recognition is lacking in personal, social, or institutional relationships, the people involved experience a form of injustice. Through struggles for recognition, they can try to improve those recognition relations. Major examples here are the Civil Rights Movement in the US, that fought for the expansion of the rights of African Americans; or the more recent #MeToo movement, which meant to end sexual violence and abuse of women.
Recognition and technology
Traditionally, recognition relations are taken to exist exclusively between human persons. Even in the case of institutionalized recognition (e.g. the recognition of someone’s freedom of speech through corresponding rights), the law can be seen as an instrument or mediator of interpersonal recognition, namely by representing the views and values of the members of a society.
Like the law, technology too can be understood as an instrument or mediator of recognition relations between human individuals or groups. Take communication technologies and social media platforms; these are instruments that enable people to express their love, esteem, and respect to others. At the same time, technologies can make the existence of misrecognition between certain societal groups visible. This is for instance the case when social inequalities are reflected in the training data of AI models, which in turn causes algorithmic bias.
But AI today has become much more than an instrument for communication, surveillance, research, or entertainment. With the rise of generative AI, and in particular LLM-based chatbots, the phenomenon of ‘AI companions’ has emerged. AI systems can now take on the role of social partners, which was previously reserved exclusively for fellow humans. This raises the question: Can an AI system be a giver and/or recipient of recognition in human-AI interaction?
Sycophancy in human-AI interaction
The term ‘sycophancy’ refers to insincere flattery or excessive agreeableness. Chatbots and other interactive AI systems are often accused of being sycophantic, because human-AI interaction tends to be highly positive and affirmative. This can be explained by the fact that chatbots tailor their outputs to what users most probably like to hear.
However, where critics may see probabilistic outputs that lack sincere attitudes, users may experience their affirmative interactions with AI chatbots as instances of recognition. Users also build lasting relationships with these systems, that extend beyond single interactions. In fact, marketed as ‘AI companions,’ forming enduring and intimate bonds is exactly the goal of many chatbot providers.
Intersubjectivity, reciprocity and functionality
So, what to make of experiences of recognition in human-AI interaction? While AI companions may lack the ability to act in accordance with sincere recognitive attitudes or mental states, they clearly are able to perform recognitive acts. When those recognitive acts are perceived by human users as meaningful instances of recognition, despite the lack of mental states on the part of the AI system, they can still impact the human user. Real-life impacts of recognitive acts are for instance the improvement of someone’s self-esteem, self-confidence, or general sense of identity.
Interpersonal recognition is intersubjective; both the giver and recipient of recognition are dialectically shaped by the relation. In human-AI relations, only one party has subjectivity. Hence, only the human in human-AI interaction can receive recognition. This recognition may not be genuine or sincere, but it may nevertheless be functional. Hence, to complement existing theories of recognition, and to make sense of the impact of the sycophantic behavior of AI companions, I propose the introduction of a new concept or category within recognition theory: functional recognition.
For more on this topic, see:
Waelen, R. (Ed.) (2026). Normative Relations of Recognition in the Age of AI. On Critical Theory and Technology. Springer.
Leave a Reply