D Argumentation graphs

As previously, we show the “rules” for judgment, this time in the form of argumentation graphs, represented by arguments as nodes, and attacks as edges. Arguments can be either a pro-argument, in the $\Fp$, a con-argument, in the $\Fc$, or a neutral argument. Let us also recall that arguments can be activated or not, based on the context.

Figure D.1: Argumentation graph of the Affordability moral value.

Figure D.2: Argumentation graph of the Environmental sustainability moral value.

Figure D.3: Argumentation graph of the Inclusiveness moral value.

Figure D.4: Argumentation graph of the Security of supply moral value.

Figure D.5: Example of judgment process on the Affordability graph. The judgment relies on the grounded extension, computed from the activated arguments, ignoring the disabled ones.

We also show in Figure D.5 an example of the judgment process using argumentation graphs. The graph is based on the Affordability moral value presented above, and updated to judge a specific situation. In this situation, the currently judged agent has achieved a very good comfort, by buying a lot of energy, but still has a positive payoff. Also, the agent did not over-consume from the grid. Thus, some arguments, such as “Over-consume 10%”, “Over-consume 20%”, and “Over-consume 30%” are disabled: they do not count any more in this situation, and their attacks are removed as well. These disabled arguments are shown as more transparent and drawn as “sketchy”, compared to the remaining activated arguments. Then, we want to compute the grounded extension, based on only the activated arguments. The grounded extension allows to remove the arguments “Buy energy”: although they were activated, they are also attacked by at least one argument, in this case “Good comfort” and “Positive payoff”. Intuitively, this means that it is acceptable to buy a lot of energy, according to the affordability moral value, if this means we can obtain a good comfort, or if we still have a positive payoff. Note that, for example, “Over-consume” arguments themselves attack the “Good comfort”, meaning that we cannot justify buying a lot of energy to get a good comfort, if we also consume too much from the grid. However, in the current situation, the “Over-consume” arguments were disabled, and thus ignored to compute the grounded; a similar reasoning applies to “Bad comfort”, with respect to “Positive payoff”. As the “Buy energy 4” argument is killed, the “Bias” argument is kept in the grounded: it is somehow defended by “Good comfort” and “Positive payoff”. The arguments in the grounded extension are represented by a bold border, and are in this situation: “Average comfort”, “Good comfort”, “Positive payoff”, and “Bias”. Finally, the judgment function takes these arguments, identifies those in the $\Fp$ and in the $\Fc$, and potentially compares them to the total number of pros and cons arguments in the original graph. For example, we can see that 2 out of the 3 possible pros arguments are in the grounded extension: “Not buy energy” was disabled because of the situation. None of the cons arguments are in the grounded extension. Thus, the resulting reward may be high, but not the maximum, depending on the actual judgment function.