Skip to main content
. Author manuscript; available in PMC: 2022 Nov 29.
Published in final edited form as: Proc SIGCHI Conf Hum Factor Comput Syst. 2022 Apr 29;2022:440. doi: 10.1145/3491102.3501886

Figure 7:

Figure 7:

Change in q-values over 200 training episodes for the goal “Choose lean proteins,” when only two foods are mentioned and one is a protein. In this state, even though there is an ambiguous protein to as a follow-up “what kind” question about, the q-values reward asking “search” questions like “what else is in <x_food_item>”, demonstrating a balance between valuing “search” and “drill-down” question types.