Managing dialogue: How information availability affects collaborative reference production

Explicit feedback from users attenuates memory biases in human-system dialogue

“…Specifically, the effect of acceptance was not statistically significant, failing to replicate a finding reported in several studies on human-human dialogue (Knutsen & Le Bigot, 2012Knutsen et al, in press) and preventing us from concluding that the acceptance bias also occurs in human-system dialogue. This is potentially due to the fact that in previous human-human dialogue studies, dialogue partners were free to choose how they accepted the information they were presented (i.e., they could choose between accepting a piece of information through verbatim repetition, implicitly, or through any other mean).…”

“…This raises a number of theoretical questions. In particular, research on human-human dialogue suggests that presentation and acceptance affect not only reference accessibility in memory after the end of an interaction, but also reuse throughout that interaction (self-presented and/or references accepted through verbatim repetition are more likely to be reused than any other reference; Knutsen & Le Bigot, 2012Knutsen et al, in press). Future research should seek to determine whether this is also the case in human-system dialogue.…”

“…Indeed, after the end of an interaction, each partner remembers the information that he or she presented him-or herself better than the information presented by his or her partners; information accepted through verbatim repetition is also remembered better than information accepted implicitly (regardless of whether the acceptance was self-or partner-produced). Such memory biases have important consequences for subsequent partneradaptation, as readily accessible information is more likely to be reused in the remainder of the interaction (Knutsen & Le Bigot, 2012Knutsen et al, in press). …”

“…Twenty-four restaurant categories (which corresponded to 24 restaurant categories found on the French website http://www.linternaute.com), 24 prices (going from five euros to 62 euros) and 24 locations (which were randomly selected from the maps used by Knutsen & Le Bigot, 2012 what's more, the map used in the current study represented the same area as the map used by Knutsen & Le Bigot, 2012, Experiment 1) were randomly associated to create twenty-four "category-location-price" items included in the lists used by the participants to interact with the simulated dialogue system (see Appendix A for an example).…”

“…It featured a total of 40 landmarks, 24 of which were the landmarks featured on the list and the remaining 16 were distractors (just like the 24 target landmarks used in this study, these 16 distractors were randomly taken from the materials used by Knutsen & Le Bigot, 2012. Two different versions of the map were created and printed on A4 sheets; the (random) position of each landmark varied across maps, ensuring that the participants' performance on the memory test was not due to the location of the landmarks on the map (see Appendix B for an example of a map).…”