Public Interest Communication (PIC) seeks to deliver messages to the public through communication strategies that align with ethical and social values. Despite its social relevance, the field currently lacks formal tool-supported methods for designing value-sensitive campaigns. This paper presents AVAPla, a web-based tool developed within the EPICA project to assist in planning PIC campaigns through computational argumentation. AVAPla enables users to construct arguments that consider both audience characteristics and underlying values, ofering visual and analytical features to assess their potential impact and efectiveness. A use case is provided to demonstrate how the tool can support more targeted and persuasive public interest communication.
Public Interest Communication (PIC) is an emerging field focused on how to efectively communicate
about social issues in ways that foster public engagement and debate. It draws on disciplines such
as public relations, political communication, advocacy, and activism, all of which are concerned with
the ethical design and governance of public communication campaigns [
AVAPla is a computational tool designed to support the structured design, selection, and organization
of arguments to achieve a communicative goal. It is specifically created to help design public interest
campaigns that are tailored to value-sensitive audiences. While web-based interfaces for creating
argument maps as graphs are available [
This paper introduces AVAPla, presents its features in Section 2 and reports a case study in Section 3. Future developments are reported in the concluding section.
The tool is provided as a web interface1, designed to support the modelling and analysis of public interest communication through computational argumentation techniques. Each argument is associated with a vector that captures its afinity for diferent values that influence public perception of the communication. By exploring these value arrays, the tool ofers insights into how diferent argumentative strategies may resonate with specific audience segments. User interaction unfolds in three main steps: campaign definition, visualisation of various computed metrics, and analysis through predetermined goals.
The campaign is modelled starting from a Value-based Argumentation Framework ⟨, →, pos⟩, defined
in [
∑︁ = 1 and =1 ∀ ≤ , ≥ 0.
Each audience has its own preferences over the values. These are represented by a function asv : → , which assigns to each audience a vector. The -th entry of asv() shows how important value is to audience .
In the tool, the campaign structure is defined through a JSON schema 2 that formalises the components of a framework ⟨, →, pos⟩ enriched with values and audiences. The schema requires the specification of seven main elements: positiveArguments, arguments, attacks, values, audiences, argumentValues, and audienceValues. Arguments are divided into two sets: positiveArguments, which contain the core arguments pos, and a general arguments set, which includes all other relevant arguments from ∖ pos. Each argument includes a unique identifier and an array of sentences representing its textual content. The attacks component defines the binary relation →, where each attack is specified by an identifier of the source and target argument. The values array lists value identifiers used in the campaign, then associated with arguments via argumentValues, which, like the val function, links each argument to one or more values, using a weight between 0 and 1 to represent the strength of the connection. Similarly, the audiences array defines each audience ∈ by a name and a weight , while audienceValues shows how much each audience cares about
2Schema available at: https://epica.dmi.unipg.it/tool/script/schema.js. each value, using weights to express these preferences, efectively implementing the function asv. To define a public interest campaign, users may choose from one of the available input methods: uploading or pasting a JSON file, or using a guided input form.
The second step allows users to explore the campaign by computing and visualising in tables some of
the measures proposed by [
Impact measure table For each audience ∈ and argument ∈ , AVAPla computes the impact function
1 ‖‖ = √ ‖asv() ⊙ val()‖ that measures the influence of on , based on how well the values promoted by the argument align with those prioritised by the audience.
Defeat relation table To include values and audience preferences in the framework, AVAPla computes a defeat relation ↠ for each audience ∈ . Given two arguments , ∈ , we use this definition of defeat
↠ ⇐⇒ ( → ∧ ‖‖ ≥ ‖ ‖).
This ensures that an argument can only defeat another if it attacks it and has at least as much impact on the audience.
Acceptability table AVAPla checks whether an argument ∈ convinces a specific audience by computing the grounded semantics [8] on ⟨, ↠ ⟩. The result is con(), which is true if and only if the argument is accepted in the grounded extension. The tool computes and shows the value of con() for every argument and audience.
Campaign graph The tool also provides an interactive graph to explore the campaign and facilitate the facilitate interpretation of the results. The graph shows the campaign from the perspective of a selected audience, which can be chosen and toggled by the user via a dropdown menu. Visual styles in the graph help distinguish between elements from positiveArguments and arguments (the nodes), attacks and the computed defeat relation ↠ (the edges), and the arguments accepted as convincing by con().
Users can interact with the graph in various ways to add arguments, create attacks, delete elements, and move arguments around. Each interaction triggers a recomputation of all the measures from step 2, ensuring that the graph remains consistent with the campaign representation. For example, when an attack is added, the system checks whether it qualifies as a defeat and updates its visual style accordingly. When adding a new argument node to the campaign graph, the user is asked to provide some information to define the argument. These include: an ID, the type (either positiveArguments or arguments), a natural language sentence expressing the argument’s core idea, and a set of weights associated with the diferent values. This ensures that the argument is fully integrated into the framework and enables a meaningful computation of its impact and interactions with other arguments. In addition to these interactions, a tooltip appears when hovering over an argument in the graph, displaying information including the argument’s computed impact for the selected audience.
The final step allows users to analyse the campaign through two distinct goal functions. Overall Efectiveness selects the argument ∈ pos that achieves the highest impact across all audiences, weighted by . For each positive argument, the tool computes the quantity
∑︁ · ‖ ‖.
=1 On the other hand, the Convinced People goal selects ∈ pos that convinces the largest share of the audience. In this case, the tool computes the quantity
∑︁ · [con()] (with [ ] = 1 if is true, 0 otherwise).
=1 Once the analysis is computed, a table is shown with the weighted scores of all positive arguments according to each goal function.
In this section, we present a simple illustrative use case to show the functionalities and analytical capabilities of the AVAPla tool in the context of planning a Public Interest Communication (PIC) campaign. A government agency commissions a team of communication experts to design a campaign — referred to as MoreGreens — aimed at encouraging the consumption of fruits and vegetables. The target audience includes two particularly resistant demographic groups: individuals under 25 years old (hereafter, young people) and individuals over 60 (hereafter, older adults). Survey data previously collected on these two groups reveals distinct value orientations. Young people tend to pursue lifestyles centred on pleasure, and are thus motivated by the value of hedonism. In contrast, older adults focus on maintaining their health and physical well-being, making security their guiding value.
Although these two values are not inherently incompatible and can, in theory, be addressed by the same argument, in this scenario, the communication experts have identified seven arguments for the MoreGreens campaign, each of which exclusively appeals to a single value. Of these, two are strongly aligned with hedonism and three with security. The remaining two arguments act as counterarguments: one challenges two pro-hedonism arguments, and the other a pro-security argument. Both counterarguments strongly promote hedonism or security, respectively.
In the first case, focusing on the young people audience, the counterargument is able - at least in one instance - to defeat the corresponding positive argument, thereby reducing its persuasive impact on the target group (see Figure 1).
If the campaign’s objective is to maximise the persuasiveness of selected arguments across both audiences —each guided by diferent value systems— it becomes essential to identify arguments that resonate with the largest group of people. These are determined through AVAPla’s Convinced People function. Naturally, arguments that are defeated (i.e., successfully countered) are excluded from this selection process. AVAPla thus supports the identification of the most promising arguments on which an efective campaign can be built.
In this paper, we introduced AVAPla, a tool designed to support the planning phase of public interest campaigns, providing insights into argument efectiveness based on measures from computational argumentation. We believe that PIC could benefit from dynamic tools that support the efective planning of value-sensitive campaigns. At the same time, we believe that computational argumentation should investigate interesting aspects — such as the dynamic construction of the audience — that emerge as relevant in the actual practice of PIC but would be considered marginal in a purely formal approach.
As a future development, we plan to integrate AVAPla with a natural language processing pipeline capable of automatically extracting core arguments from a set of documents, identifying attack relations, and assigning each argument a graded value based on the finite set of universal values. Within this broader framework, we also aim to model the perceived credibility of the source of the arguments — as judged by the target audiences of the PIC campaign — as an additional parameter that can influence the overall efectiveness of the positive arguments.
Carlo Taticchi is a member of the INdAM Research group GNCS and of Consorzio CINI. This work has been partially supported by INdAM - GNCS Project, codice CUP_E53C24001950001 and MUR project PRIN 2022TXPK39 - PNRR M4.C2.1.1. “Empowering Public Interest Communication with Argumentation (EPICA)” CUP H53D23003660006, funded by the European Union - Next Generation EU, Missione 4 Componente 1.
[8] P. M. Dung, On the acceptability of arguments and its fundamental role in nonmonotonic reasoning and logic programming, in: R. Bajcsy (Ed.), Proceedings of the 13th International Joint Conference on Artificial Intelligence. Chambéry, France, August 28 - September 3, 1993, Morgan Kaufmann, 1993, pp. 852–859. URL: http://ijcai.org/Proceedings/93-2/Papers/003.pdf.