Emotional AI in customer service pleased only some shoppers, experiments with 677 people find

For communal consumers, those who expect a caring, relationship-style bond with a brand, high emotional expression raised satisfaction and continuance intention compared with low expression. For exchange consumers, who treat dealings as transactions, the effect was non-significant.

The effect worked through expectation confirmation, meaning satisfaction rose when the AI's emotional behaviour matched what customers expected. It appeared for communal consumers only when they perceived the AI as having an experiential mind, so a blanket approach to emotional AI would not benefit all customers.

The study examined emotion-expressing AI in customer service, where earlier outcomes have been mixed and dependent on context. It tested the joint effect of how emotional the AI appeared and the type of relationship consumers expect with a brand, communal or exchange.

The outcomes measured were consumer satisfaction and continuance intention. A significant interaction showed emotional expression did not work the same way for everyone: it helped communal consumers and made no significant difference for exchange consumers.

The authors identify expectation confirmation as the mechanism and AI mind perception as the boundary condition. Their practical advice is to calibrate rather than standardize emotional expression by relationship norms, and to raise customers' perception of the AI's experiential capacity through interaction design and user education.