Welcome to the EPICA Tool!
This tool implements a framework for measuring the effectiveness of public interest communication by analysing arguments according to multiple values. Arguments are represented as vectors that reflect their relationship to different factors influencing audience perception and response. By examining these value dimensions, the tool provides insights into the impact of communication strategies across various audience groups. It is designed to support research and analysis of public campaigns.
Step 1: Campain Definition
Campaign Definition
The campaign is modelled using a Value-based Argumentation Framework, represented as a triple \( \langle A, \to, A^{\text{pos}} \rangle \). Here, \( A \) is a set of arguments, \( \to \) is a binary attack relation between arguments, and \( A^{\text{pos}} \) is a subset of A representing the core arguments expressing the campaign’s goals. To enhance the framework, a value structure and audience model are introduced.
Arguments can appeal to multiple values simultaneously, each to a different degree. To represent this, the value space is defined as \( V = [0,1]^n \), where each dimension corresponds to a specific value. A value function \( \text{val} : A \to V \) assigns to each argument a a vector indicating its degree of association with each value.
The set of audiences is defined as \( I = \{1, 2, \dots, k\} \), where k is the number of distinct audience groups. Each audience i is associated with a weight \( p_i \) representing the proportion of the overall population sharing similar values. These weights satisfy the conditions: \[ \sum_{i=1}^k p_i = 1 \quad \text{and} \quad \forall i \leq k, p_i \geq 0 \] Each audience group i has individual preferences over the values, represented by an audience-specific value function \( \text{asv} : I \to V \). This function assigns to each audience i a vector where the j-th entry indicates the importance that audience i gives to value j.
Define the public interest campaign using one of the methods proposed below.
Choose an option
Upload JSON file
Step 2: Visualisation
Explore argument impact, defeat relations, grounded acceptability, and the campaign graph.
Impact measure
For every audience \( i \leq k \), an impact function \( \| \cdot \|_i : A \to [0,1] \) is introduced. This function quantifies the influence that an argument \( a \in A \) has on audience \( i \), based on the alignment between the values promoted by the argument and the values considered important by that audience.
The impact is computed using the Euclidean norm of the component-wise (Hadamard) product between the argument’s value vector and the audience’s value preference vector. Formally, the impact is given by: \[ \|a\|_i = \frac{1}{\sqrt{n}} \left\| \text{asv}(i) \odot \text{val}(a) \right\| \] This ensures that an argument relying on values deemed important by an audience will receive a higher impact score for that audience. The following table shows the impact \( \|a\|_i \) for each \( a \in A \) and audience \( i \in I \).
Defeat relation
Given two arguments \( a,b \in A \) and an audience \( i \in I \), the defeat relation is defined as \( a \twoheadrightarrow_i b \iff (a \rightarrow b \land \|a\|_i \geq \|b\|_i) \). This notion captures the idea that an argument can only defeat another when it is equally or more persuasive, based on the value priorities of the given audience. The following table shows the defeat relation \( \twoheadrightarrow_i \) for every audience \( i \in I \).
Acceptability (grounded semantics)
To determine whether an argument \( a \in A \) will convince a specific audience \( i \), denoted as \( \text{con}_i(a) \), the following algorithm is used:
- Consider the argumentation framework \( \langle A, \twoheadrightarrow_i \rangle \)
- Add undefeated arguments to the grounded extension \( \varepsilon_{\text{GR}}(A) \)
- Remove from \( A \) all arguments defeated by elements of \( \varepsilon_{\text{GR}}(A) \)
- If the modified framework contains undefeated arguments, go to step 2, otherwise terminate
- Define: \( \text{con}_i(a) \iff a \in \varepsilon_{\text{GR}}(A) \)
The following table shows \( \text{con}_i(a) \) for each audience \( i \in I \).
Campaign graph
Below is shown the graph related to the described campaign. The following interactions are possible.
- Add an argument: click anywhere on the canvas
- Add an attack: first, click on the source argument, then click on the target argument
- Delete an argument or attack: click on the item, then press Backspace or Delete
- Move an argument: hold down the Shift key, then drag the argument to a new position
ID:
Type:
Sentence:
Impact: ()
Step 3: Analisys
Goals represent the task of identifying the most effective positive arguments in the campaign to persuade the audience.
Overall Effectivenes: find the argument \( a \in A^{\text{pos}} \) for which the following quantity is maximal: \[ \sum_{i=1}^{k} p_i \cdot \|a\|_i \]
Convinced People: find the argument \( a \in A^{\text{pos}} \) for which the following quantity is maximal: \[ \sum_{i=1}^{k} p_i \cdot \chi(\text{con}_i(a)) \] where \[ \chi(\varphi) = \begin{cases} 1 & \text{if } \varphi \text{ is true} \\ 0 & \text{if } \varphi \text{ is false} \end{cases} \] and \( \text{con}_i(a) \) is true if and only if argument \( a \) is able to convince audience \( i \).