AI assistants: should they have a personality?
No-code tools make it easy to give an AI assistant a name, a tone and an identity. For a manager preparing its deployment, the question is specific: does this presentation make the assistant easier to use, or does it mainly increase trust without improving the service delivered?
What the literature establishes
Clifford Nass and Youngme Moon (2000, Journal of Social Issues) analyse experiments showing that people apply social rules to computers that are normally reserved for human interactions. These responses do not require users to believe that the machine is actually human. This distinction matters when designing an assistant: knowing that one is using software does not necessarily neutralise the effect of its presentation.
When presentation changes trust
Adam Waytz, Joy Heafner and Nicholas Epley (2014, Journal of Experimental Social Psychology) show, in a simulated driving experiment, that an anthropomorphic presentation of an autonomous vehicle increases trust in it. Their design combines features including a name, a voice and a gender, and the authors link the effect to the attribution of a mind to the machine. The finding concerns a vehicle, not a conversational assistant, but it offers a relevant mechanism to examine: trust can vary with a system’s human-like cues.
The premature conclusion
One might conclude that every assistant should be made more human-like to encourage adoption. Yet these studies do not show that greater trust improves decision quality or that it better matches the system’s capabilities. The managerial task is therefore not to maximise the interface’s appeal, but to check that its presentation does not make limited assistance look like expertise. (our executive and employee training programmes)
What these findings cannot settle
These articles predate assistants based on large language models and do not evaluate today’s no-code platforms. The vehicle experiment also does not allow the observed effect to be attributed separately to the name, voice or gender. Nor does it establish that these cues have the same effects in French and German: in a bilingual organisation, that equivalence still needs testing.
A practical check in Fribourg
As part of the programme « Créer ses assistants et agents IA sans coder », a Fribourg pilot could bring together an agri-food company, a bilingual industrial team and participants from the local university community to work on non-sensitive internal requests. Randomly assigning participants to either a functional interface or one featuring a first name and an avatar, while keeping responses strictly identical and balancing the French and German versions, would help isolate the effect of this presentation. The main measure would be the proportion of responses accepted without changes despite a known limitation, built into the test cases and explained during the debriefing. Comparing this proportion between interfaces, separately for each language, would provide a verifiable indicator of presentation-driven acceptance rather than a simple satisfaction score. To go further: explore the Building AI Assistants and Agents Without Coding training in Fribourg, or browse our executive and employee training programmes in Switzerland.
In pictures: Building AI Assistants and Agents Without Coding in Fribourg
- building ai assistants and agents without coding
- Fribourg
- research
- training fribourg