Artificial Intelligence (AI) is expected to deliver many benefits for digital citizen participation, e.g., by allowing citizens to comprehend larger amounts of data or providing balance to the diference in resources between citizens and governments. However, using AI in citizen participation is not a straightforward task. Its inherent complexity, opaque nature, and power to influence public discourse plague AI developments with a substantial number of risks. Making matters worse, the currently nascent literature in this area provides a limited understanding of the implications of AI for citizen participation. Our BeCoDigital research project aims to shed some light on the use of AI in citizen participation by providing a roadmap on how to use AI within this context. Our roadmap provides insights and recommendations around three aspects. WHAT are the current applications of AI in citizen participation, WHY should AI be used in this context according to citizens and practitioners, and HOW (i.e., which principles to follow) to implement AI solutions in citizen participation.
The diversity of AI-driven techniques provides a rich variety of solutions that could be transferred and exploited in citizen participation. However, the scattered literature between fields makes it hard to evaluate the current state-of-the-art, the level of compatibility between these approaches, and their relevance for citizen participation. In fact, AI is a broad term associated to several technologies that do not necessarily share the same characteristics and social impacts. Thus, before implementing AI systems in citizen participation, there is a clear need to develop a theoretical foundation that shows how to operationalize these emerging technologies in citizen participation. We achieve this by analyzing the literature at a more abstract level. We define a typology common to all the AI-enhanced solutions related to digital participation platforms. The typology is built along two dimensions. First, the role of AI, that can either take over tasks for humans (automation) or work collaboratively with them (augmentation). Second, the scope of the AI, that can work at the level of a single citizen (individual), of the interaction between citizens (peer-to-peer), or of the citizen participation initiative globally (collective). Thus, our typology describes 6 types of AI solutions for citizen participation summarized in Table 1. Beyond its descriptive contribution, the typology can also be used to evaluate future avenues and challenges in the use of AI in citizen participation.
.ttaom tPPhrreooccineesdsssiivinnidgguitnaoflootlarsmskastoiofnththeaptafraticciilpitaantetss u during the participatory process. A t Individualized feedback . en AI interaction to improve the perform mance, motivation or knowledge of g uA individuals.
According to Technological Frames theory, misalignment between the expectations of stakeholders
involved in a technological development can lead to a negative impact on acceptance and
counterproductive efects. This issue is all the more pressing in the case of AI use in citizen participation,
since previous works have reported diverging views on citizen participation and AI across citizens
and government practitioners. For AI to be used in citizen participation in a way that is aligned with
expectations and realizes its potential, it is essential to investigate how citizens and practitioners perceive
AI use in this context, and to identify and solve any misalignment. We achieve this by collecting the
discourse related to AI use in citizen participation and applying the Q-methodology to collect, analyze,
and compare the subjective viewpoints of citizens and practitioners.
Current AI solutions focus mainly on technical challenges, neglecting their social impact and not fully
exploiting the potential of AI to empower citizens. We investigate how to design digital participation
platforms that integrate technical AI solutions while considering the social context in which they are
implemented. Using Collective Intelligence as kernel theory we generate design principles for the
development of a socio-technically aware AI architecture. We validated them with AI and citizen
participation experts. The principles suggest (
The research pertaining to these results received financial aid from the Belgian Science Policy Ofice according to the agreement of subsidy no. [B2/223/P3/BeCoDigital].