Backchannels: Quantity, Type and Timing Matters

In a perception experiment, we systematically varied the quantity, type and timing of backchannels. Participants viewed stimuli of a real speaker side-by-side with an animated listener and rated how human-like they perceived the latter's backchannel behavior. In addition, we obtained measures of appropriateness and optionality for each backchannel from key strokes. This approach allowed us to analyze the influence of each of the factors on entire fragments and on individual backchannels. The originally performed type and timing of a backchannel appeared to be more human-like, compared to a switched type or random timing. In addition, we found that nods are more often appropriate than vocalizations. For quantity, too few or too many backchannels per minute appeared to reduce the quality of the behavior. These findings are important for the design of algorithms for the automatic generation of backchannel behavior for artificial listeners.