almost complete definition
almost complete. . This means that the actions of all synthetic participants, i.e., which comments they have given feedback to, and which values, are specified, except for the controlled participant. We then generate two completions of this state: One includes the actions of the controlled participant following the ’always truthful’ strategy, the other one includes his actions when following the reference strategy, e.g., ’always untruthful’. We compute the utility of the controlled participant in these two cases. We refer to such a sequence of steps as simulation run. If the utility of ’always truthful’ is higher (not lower) than the other one, this is a good sign. We repeat these simulation runs – which are random processes, as we will explain right away, i.e., the new state will most likely be different from the previous one – and again compare the utilities. We keep doing this until there is some statistical significance that the controlled participant can expect a higher utility from one of the strategies.