In AI we trust : A qualitative investigation into the concept of trust in human-AI relations and its changing nature

Sammanfattning

With the increasing use of AI and its growing influence, the need to understand how humans build trust in technology and communicative agents, such as AI chatbots, is becoming increasingly important. Through qualitative interviews with ten different participants with varying backgrounds, the thesis investigates three questions: How do people perceive trust concerning AI-chatbots? What do people perceive that they trust or distrust when using AI-chatbots; and how, and if, is trust formed and understood through people’s interactions with AI-chatbots? Using trust theories from Niklas Luhmann and Piotr Sztompka and Actor-Network-Theory, the research discloses the complexity of trust as a layered, negotiated, and contradictory concept through its cognitive, emotional, and cultural dimensions. The results showed interesting tendencies where participants, exercising cognitive trust based on reliability and functionality, could at times develop emotional trust ties to the machines based on personalization, simulating social cues and language, despite being aware of the AI’s nature. Distrust emerged in connection with the intentions of developers and companies, lack of transparency, data harvesting, and manipulation based on the chatbots' behavior and design. This, however, did not stop users from accepting receding control and autonomy to the AI, showing the continuous negotiation and redirection of trust in AI as a multifaceted and layered process.