Trusting me, trusting you: How researchers are redesigning relationships between humans and intelligent machines

University of Manchester researchers are investigating how intelligent machines can communicate with people in ways that foster trust in workplaces, hospitals and homes. Their work contrasts with the 1951 Ferranti Mark I era, when programmers had to translate instructions into machine-oriented symbols. The current goal is for machines to better recognize human intentions, changes in demeanor and social expectations.

The researchers report that trust in robots depends both on whether a robot delivers what it claims and on how it handles situations, particularly failures. A robot that completes a task while missing ordinary social cues may be trusted less than one that fails but acknowledges the error in a human-like manner. In a study on voice and trust, participants showed more confidence working with a robot using an AI-generated voice designed to sound human, although they later said they preferred a standard robotic voice.

The team is also developing Computational Trust models, which enable robots to judge how dependable human collaborators may be. Such estimates can help a robot anticipate a partner’s actions and possible errors. This work connects to artificial Theory of Mind: giving machines some ability to infer that others have distinct knowledge, needs, intentions and goals.

In one study, autonomous robot pairs had different personalities, priorities and hidden needs. They used their own experiences to interpret each other’s behavior and refined their estimates through interaction. Cooperation was strongest when at least one robot prioritized the other’s needs; reasoning about another agent’s state alone was insufficient.

Manchester’s Emma system was trained with facial expressions and text to recognize emotional states and produce empathetic responses. Compared with a leading commercial language model, its replies were judged more humanlike and appropriate. The researchers argue that transparency, honesty and understanding human needs will be central to dependable human-machine relationships.