It is beneficial to identify and begin treatment of neurocognitive disorders of the elderly as early as possible. In order to help diagnose these disorders, social assistive robots are promising technologies to assist psychologists. To be accepted by the elderly, the robot behaviours must be close enough to the fundamental competences of the psychologists in order not to confuse the patient. This pilot study aims (1) to design a social assistive robot capable of performing a memory evaluation test, (2) to gather opinions on the robot's acceptability with an innovative method (persona) and (3) to identify robot behavioural improvements. We used the "persona methodology" for this pilot study. A panel of students playing the role of a "persona" performed the memory test called RL/RI16, by interacting with the social robot Pepper and then were interviewed about their experience. The robot plays the psychologist role. The interviews and videos analysis showed that the robot is not yet well accepted but the analysis results gave interesting leads to continue.