Argumentation-based Dialogue for choosing a car engine

This is an implementation of an argumentation model with 2 agents arguing between types of car engines. Each engine has five properties : cost, consumption, durability, environmental impact and degree of noise. Each agent has a list of property preferences.

Authors

Fernando Kurike Matsumoto

Victor Felipe Domingues do Amaral

Installation and execution

To install the dependencies, use pip and the requirements.txt file:

pip install -r communication/requirements.txt

To run the code with the example items and preferences:

python3 pw_argumentation.py --preset

To run the code with random items and preferences, use the --number-items=N argument. For example, for 10 random items:

python3 pw_argumentation.py --number-items=10

Additional flags supported by the script are --prob-acccept-argument, --first-agent and --top-k. These behavior of these parameters is described in the next section. For more information about the accepted flags, see:

python3 pw_argumentation.py --help

Implementation aspects

All of the model's behaviors were implemented in accordance with the instructions from the course, except for certain aspects listed below:

• The constructor of the model allows for the option of passing a CSV file path that contains the preferences of each agent and items in the format found in the data folder. If no file is passed, then the preferences and items will be randomly generated. The number_items parameter specifies the number of engines to be created. If it is not None, the preferences of each agent will also be randomly selected.

• The parameter first_agent ∈ {"Alice", "Bob"} specifies the name of the agent that will initiate the conversation.

• The function is_item_among_top_10_percent() identifies whether an item is among an agent's top 10% preferred items. To ensure that at least one item is returned when there are fewer than 10 items in total, we take in consideration the ceiling of the number of items divided by 10.

• We introduced a parameter top_k indicating to is_item_among_top_10_percent() the percentage of items to consider.

• If an agent cannot attack the other agent's latest argument, they will be convinced by the argument and accept the item with probability prob_accept_item (a parameter we introduced). Otherwise, they will try to propose the best item that hasn't been proposed yet.

Example of arguments generated

From Alice to Bob (PROPOSE) Engine8
From Bob to Alice (ASK_WHY) Engine8
From Alice to Bob (ARGUE)     Engine8 <- DURABILITY=GOOD
From Bob to Alice (ARGUE) not Engine8 <- CONSUMPTION=VERY_BAD, CONSUMPTION>DURABILITY
From Alice to Bob (ARGUE)     Engine8 <- DURABILITY=GOOD, DURABILITY>CONSUMPTION
From Bob to Alice (PROPOSE) Engine5
From Alice to Bob (ASK_WHY) Engine5
From Bob to Alice (ARGUE)     Engine5 <- CONSUMPTION=VERY_GOOD
From Alice to Bob (ARGUE) not Engine5 <- DURABILITY=BAD, DURABILITY>CONSUMPTION
From Bob to Alice (ARGUE)     Engine5 <- CONSUMPTION=VERY_GOOD, CONSUMPTION>DURABILITY
From Alice to Bob (PROPOSE) Engine7
From Bob to Alice (ASK_WHY) Engine7
From Alice to Bob (ARGUE)     Engine7 <- CONSUMPTION=VERY_GOOD
From Bob to Alice (ARGUE) not Engine7 <- CONSUMPTION=VERY_BAD
From Alice to Bob (PROPOSE) Engine8
From Bob to Alice (ASK_WHY) Engine8
From Alice to Bob (ARGUE)     Engine8 <- DURABILITY=GOOD
From Bob to Alice (PROPOSE) Engine4
From Alice to Bob (ASK_WHY) Engine4
From Bob to Alice (ARGUE)     Engine4 <- CONSUMPTION=VERY_GOOD
From Alice to Bob (ARGUE)     Engine7 <- PRODUCTION_COST=VERY_GOOD
From Bob to Alice (ARGUE) not Engine7 <- CONSUMPTION=VERY_BAD, CONSUMPTION>PRODUCTION_COST
From Alice to Bob (PROPOSE) Engine8
From Bob to Alice (ASK_WHY) Engine8
From Alice to Bob (ARGUE)     Engine8 <- DURABILITY=GOOD
From Bob to Alice (PROPOSE) Engine6
From Alice to Bob (ASK_WHY) Engine6
From Bob to Alice (ARGUE)     Engine6 <- CONSUMPTION=VERY_GOOD
From Alice to Bob (ARGUE) not Engine6 <- CONSUMPTION=BAD
From Bob to Alice (PROPOSE) Engine5
From Alice to Bob (ASK_WHY) Engine5
From Bob to Alice (ARGUE)     Engine5 <- CONSUMPTION=VERY_GOOD
From Alice to Bob (ACCEPT) Engine5
From Bob to Alice (COMMIT) Engine5
From Alice to Bob (COMMIT) Engine5
Done. Chosen product: Engine5 (No description)