Index of experiment materials

Here is a list of the experiments, in their order of appearance in the thesis, and the location of the results, input and control files that corresponds with them. A brief description of the experiment procedure is given for each.

• Chapter 4: Risking misinterpretation This experiment
  corresponds with the spanner_demos directory. Two input files were used in this experiment. config.pl was used where no clarification dialogue was desired in the game tree. config_clarify.pl differed in that plan rules for the clarification subdialogue were added. The file test_harness.pl has a procedure that runs the planner for each of the demonstrations. The data generated in the experiment was written to the data subdirectory. There is a set of gnuplot scripts in this directory which generate plots from the data.

• Chapter 4: Goal introduction problem This experiment corresponds with the window_seat_demos directory. Two input files were used in this experiment, config.pl and config2.pl. The difference between these is that the passenger who wanted a window seat but didn't get one has a lower reward in the second configuration. The file
  test_harness.pl has a procedure that runs the planner for each of the demonstrations. The data generated in the experiment was written to the data subdirectory. There is a set of gnuplot scripts in this directory which generate plots from the data.

• Chapter 4: Belief model acquisition This experiment corresponds with the window_seat_acquisition directory. The input file config.pl is similar to that used in the goal introduction experiment. The experiment was controlled using the dialogue manager program in the file domain_planner/dialogue_manager.pl. Data was collected by running the Prolog interpreter in debugging mode and examining the belief state of the agent.

• Chapter 5: Telling and proposing This experiment corresponds with the negotiation_examples/propose directory. The input file was config.pl. The test procedure in the file test_harness.pl was used to generate and evaluate the game tree. Different sets of acts were enabled and disabled to show the difference in utility they make. This was done by commenting out lines in the file
  negotiation_planner/critical_points.pl.

• Chapter 5: Holding the floor This experiment corresponds with the negotiation_examples/ask_beat_tell directory. The input file was config.pl. The file test_harness.pl was used to create two sets of results. First the value of bel(have-eggs) was set to 0.1 at level 3 for one set of results. For the second set, it was set to 0.8. These resultsets have indicative filenames in the data directory. There is a gnuplot script to generate the plots.

• Chapter 5: Request and propose This experiment corresponds with the negotiation_examples/auto_ask_got_fruit directory. The input file was config.pl. The file test_harness.pl was used to create the set of results. There is a gnuplot script to generate the plots.

• Chapter 6: Misconception This experiment corresponds with the misconception directory. The input file was config.pl. The file test_harness.pl was used to consult the input file and call the planner. Different belief revision strategies were tried by modifying the belief revision code. These modifications were thereafter removed.