Bearing in mind the ongoing hype about conversational agents associated with the rapid technological improvements and the excellent business opportunities, it is of utmost importance to incorporate all research knowledge into the development to create a successful user experience. This comprises both functional capabilities of a conversational agent’s dialog system as well as an agent’s specific design characteristics and features.
Our research team analyzed the underlying concepts of a chatbot’s design features and integrated them into a taxonomy. The taxonomy distinguishes social cues rested upon commonalities found in the means of creating them. This results in four social cue categories – the verbal, visual, auditory and invisible cues – which are based on a classical classification of the human communication system. Then we further subdevided them into ten subcategories.
Content cues refer to the strict and literal meaning of a message (i.e., what is said) (e.g., greetings, thanking, self-disclosure).
Style cues refer to the meaningful deployment of language variation in a message (i.e., how something is said) (e.g., abbreviation, formality).
Kinesic cues refer to all body movements of the agent (e.g., gestures, facial expression).
Proxemic cues refer to the role of space, distance, and territory in communication (e.g., background, conv. distance).
Agent appearance cues refer to an agent’s graphical representation (e.g., age clothing).
Computer-mediated-communication (CMC) cues refer to visual elements that can augment or modify the meaning of a text-based message (e.g., emoticons, typeface).
Voice qualities refer to permanent and adjustable characteristics of speech (e.g., pitch range, volume, tempo).
Vocalizations refers to nonlinguistic vocal sounds or noises (e.g., laugh, vocal segregate, yawn).
Chronemic cues refer to the role of time and timing in communication (e.g., response time, first turn).
Haptic cues refer to tactile sensations on the user’s body (e.g., tactile touch, temperature).