Periodic Reporting for period 1 - TITAN (AI for Citizen Intelligent Coaching against Disinformation)
Período documentado: 2022-09-01 hasta 2023-08-31
Disinformation, also known as "fake news" or intentionally false information disseminated with the aim of deceiving, poses a significant threat to democracy. In an age where information is readily available and easily spread through digital platforms, the uncontrolled dissemination of false information can have serious repercussions. This includes generating confusion, eroding trust in institutions, and fostering societal polarization. In this context, disinformation can undermine the foundations of democracy by distorting public perceptions, influencing political decision-making, and diminishing citizens' ability to make informed choices.
Social networks (e.g. X - formerly Twitter, Instagram, YouTube, TikTok, Facebook...) enable the rapid exchange of information through numerous channels and diverse sources, including unreliable ones outside mainstream media that prioritize virality over accuracy. Given the existing issues with disinformation in our digital society, Generative AI tools, such as ChatGPT and DALL-E, have the potential to exacerbate the problem by producing a vast amount of novel content at an even greater speed.
The TITAN project aims to combat disinformation by enhancing users' critical thinking skills using the Socratic method. The project seeks to develop an AI-based assistant, presented as an intelligent chatbot, which assists individuals in assessing the authenticity of a statement through personalized dialogues. The chatbot's dialogue strategy is designed to emulate the Socratic method of inquiry, customized to users' levels of critical thinking skills and potential disinformation indicators in the content at hand. The chatbot will also encourage users to utilize fact-checking tools and processes and provide access to relevant micro-lessons. It will offer an evaluation of how the web propagation of specific items may impact affected communities.
Through the use of the Socratic method, the chatbot aims to train and enhance users' critical thinking skills when confronted with disinformation. While it is not intended to be a fact-checking tool, the chatbot will serve as an intelligent coach to help users become proficient in a robust verification process.
Addressing disinformation is a multifaceted challenge, requiring the involvement of multidisciplinary teams and transdisciplinary research. This is evident in the inclusion of social sciences and humanities partners within the TITAN consortium. Contributions from the social sciences and humanities have been instrumental in laying the groundwork for the TITAN socio-technical framework, offering insights into the social, political, psychological, philosophical, cognitive, legal, and ethical factors relevant to the challenge of disinformation in an increasingly AI-driven society.
Social networks (e.g. X - formerly Twitter, Instagram, YouTube, TikTok, Facebook...) enable the rapid exchange of information through numerous channels and diverse sources, including unreliable ones outside mainstream media that prioritize virality over accuracy. Given the existing issues with disinformation in our digital society, Generative AI tools, such as ChatGPT and DALL-E, have the potential to exacerbate the problem by producing a vast amount of novel content at an even greater speed.
The TITAN project aims to combat disinformation by enhancing users' critical thinking skills using the Socratic method. The project seeks to develop an AI-based assistant, presented as an intelligent chatbot, which assists individuals in assessing the authenticity of a statement through personalized dialogues. The chatbot's dialogue strategy is designed to emulate the Socratic method of inquiry, customized to users' levels of critical thinking skills and potential disinformation indicators in the content at hand. The chatbot will also encourage users to utilize fact-checking tools and processes and provide access to relevant micro-lessons. It will offer an evaluation of how the web propagation of specific items may impact affected communities.
Through the use of the Socratic method, the chatbot aims to train and enhance users' critical thinking skills when confronted with disinformation. While it is not intended to be a fact-checking tool, the chatbot will serve as an intelligent coach to help users become proficient in a robust verification process.
Addressing disinformation is a multifaceted challenge, requiring the involvement of multidisciplinary teams and transdisciplinary research. This is evident in the inclusion of social sciences and humanities partners within the TITAN consortium. Contributions from the social sciences and humanities have been instrumental in laying the groundwork for the TITAN socio-technical framework, offering insights into the social, political, psychological, philosophical, cognitive, legal, and ethical factors relevant to the challenge of disinformation in an increasingly AI-driven society.
During the first year, the technical and scientific activities of the project have primarily focused on:
- The development of the TITAN Socio-Technical Methodological Framework Against Disinformation, with the aim of: 1)Understanding the socio-political and psychological mechanisms that contribute to the spread of disinformation; 2) Establishing the state of the art (SoA) of AI for combating disinformation; 3) Defining the legal and ethical framework for a trustworthy AI-based tool to counter disinformation.
- The formulation of ethical and legal requirements for TITAN, along with the process for assessing the legal and ethical impacts of the TITAN ecosystem.
- Determining user needs through a well-structured co-creation process to create a citizen-centered, intelligent coaching environment to combat disinformation.
- Creating the initial Critical Thinking Assessment Model, essential for personalizing dialogues with the intelligent chatbot.
- Developing AI-based disinformation signal detection models to support dialogue personalization for the intelligent chatbot.
- Establishing the first Socratic Dialogue schemes to implement a rule-based conversational agent capable of conducting personalized Socratic dialogues.
- Implementing a rule-based conversational agent that generates personalized dialogues, triggered by a news article as the basis for conversation. The dialogue personalization relies on users' critical thinking assessments and the disinformation signals they wish to investigate. This initial chatbot prototype will be utilized to assess the User Experience of participants in upcoming Living Labs, which represent the next step in the co-creation process.
- Developing the first TITAN release, scheduled for the second year of the project, utilizing an iterative development approach with a focus on iterative, incremental releases.
- The development of the TITAN Socio-Technical Methodological Framework Against Disinformation, with the aim of: 1)Understanding the socio-political and psychological mechanisms that contribute to the spread of disinformation; 2) Establishing the state of the art (SoA) of AI for combating disinformation; 3) Defining the legal and ethical framework for a trustworthy AI-based tool to counter disinformation.
- The formulation of ethical and legal requirements for TITAN, along with the process for assessing the legal and ethical impacts of the TITAN ecosystem.
- Determining user needs through a well-structured co-creation process to create a citizen-centered, intelligent coaching environment to combat disinformation.
- Creating the initial Critical Thinking Assessment Model, essential for personalizing dialogues with the intelligent chatbot.
- Developing AI-based disinformation signal detection models to support dialogue personalization for the intelligent chatbot.
- Establishing the first Socratic Dialogue schemes to implement a rule-based conversational agent capable of conducting personalized Socratic dialogues.
- Implementing a rule-based conversational agent that generates personalized dialogues, triggered by a news article as the basis for conversation. The dialogue personalization relies on users' critical thinking assessments and the disinformation signals they wish to investigate. This initial chatbot prototype will be utilized to assess the User Experience of participants in upcoming Living Labs, which represent the next step in the co-creation process.
- Developing the first TITAN release, scheduled for the second year of the project, utilizing an iterative development approach with a focus on iterative, incremental releases.
TITAN aims to advance the state of the art in the following areas:
- Advanced state-of-art in Conversational Agents: Beyond Rules, Towards Multi-signal Memory
- Privacy-preserving and Trusted Machine Learning: Trustworthy AI and Federated Learning
- Adapting (Enhancing) AI Tools: Disinformation Signals and Datasets
- Assessing the citizen’s critical thinking skills, through a novel approach based on Socratic dialogues
Even though the first year of the project was strongly focused on the groundwork for building the TITAN socio-technical framework, providing a baseline for TITAN, reflecting on social, political, psychological, philosophical, cognitive, legal and ethical drivers relevant to the need for and response to disinformation in an increasingly AI-driven society, all the results presented in the previous section are a step toward in the advancement of the state of the art indicated before.
- Advanced state-of-art in Conversational Agents: Beyond Rules, Towards Multi-signal Memory
- Privacy-preserving and Trusted Machine Learning: Trustworthy AI and Federated Learning
- Adapting (Enhancing) AI Tools: Disinformation Signals and Datasets
- Assessing the citizen’s critical thinking skills, through a novel approach based on Socratic dialogues
Even though the first year of the project was strongly focused on the groundwork for building the TITAN socio-technical framework, providing a baseline for TITAN, reflecting on social, political, psychological, philosophical, cognitive, legal and ethical drivers relevant to the need for and response to disinformation in an increasingly AI-driven society, all the results presented in the previous section are a step toward in the advancement of the state of the art indicated before.