1. Deliberation

While there is a long-history of wanting to involve the user in human-computer interaction research, with the proliferation of Participation Design (PD) and User-Centered Design (UCD), there is a lack of research on user AI systems particularly in a deliberative context. UCD emerged as computers began to be used by less expert users apart from just engineers. The change in the design approach was a need to better understand the end-user. The focus did not always have to involve the users, but users could be represented by usability experts or various theories and models of user behaviour. The principles of UCD were an early focus on users and tasks, measuring empirical data and designing in an iterative manner (Bekker & Long, 2000). While participatory design emphasises that the users be involved throughout the design process to allow them to influence the process and the resulting product. PD allows the education of various stakeholders on teaching each other about their work (Bekker & Long, 2000). 

Deliberation

Deliberation and more importantly deliberative spaces have been found by scholars to be important in the role that they play in generating ideas and information that can improve knowledge, understanding, and decision-making quality (Parkins & Mitchell, 2005). Deliberative spaces can allow citizens to discuss and debate common concerns, access a wide array of information, as well as reflect on their understanding of a variety of issues (Parkins & Mitchell, 2005). Deliberation has often been used in natural resource management and natural resource scholars focus on the importance of having various stakeholders around the table. They focus on the various stakeholders being represented at the table to allow the hearing of the concerns of the people impacted (Parkins & Mitchell, 2005). 

In recent years, deliberation has become more interesting in scholarly work as scholars have become more interested in involving the public in various discussions to get their opinions on the decisions that will impact them (Abelson et al., 2003). Deliberation allows citizens to weigh evidence on a subject, discuss and debate the various options around them and arrive to a consensus between various parties. Deliberative paradigms have been important in the healthcare sector as organisations have used various deliberative methods to engage the public in the different discussions that occur around values (Abelson et al., 2003). 

In AI

As AI has arisen in prominence, Buhmann and Fiesler (2021) have argued for a deliberative framework for responsible innovation in AI. AI Ethics is of vital importance with responsible innovation and has become to tackle the questions of transparency and explainability of systems. Buchmann and Fiesler (2021) recommend a framework that involves corporations, media and civil society in combination with each other can allow productive discourse to emerge. The authors argue for deliberation between various actors that would increase transparency, accountability, and explainability of the AI systems being used. Yeung, Howes, and Pogrebna (2020) focused on the need to consult stakeholders on the design of the system and system's potential impacts in society. Kasirzadeh and Gabriel (2022) have also argued that conversational agents can promote democratic discourse through promoting deliberation online. 

More recently, Zhang et al., (2023) proposed creating Machine models and deliberating with those models to improve and make human decision-making more fair. They proposed that machine learning model with the help of historical data can unearth successful patterns but also missteps that can involve human biases. They suppose that these models along with reflection and deliberation can allow people to have more ideas and thus improve human decision-making. In a machine learning approach, Zhang et al., (2023) focus on the importance of stakeholder input in model creation, feature selection, model training and evaluation as stakeholders can have an important say on the values impressed in the model. The researchers created a ML web tool assisted users in decision-making. They found that using ML models helped participants to generate ideas for future decision-making as well as guiding them in the participatory AI design process. The tool helped the participants realise shared interests and the intricacies of decision-making (Zhang et al., 2023).  

The tool presents an existing research case for AI systems that can assist humans in both reflection and deliberation of their values and of human-AI interaction.

On Chatbots used for deliberative purposes

Researchers have found that structured discussion may allow for better quality discussion and produce more diverse opinions within groups (Kim, Eun, Seering, & Lee, 2021). This can result in better perception of the discussion’s deliberative quality (Kim, Eun, Seering, & Lee, 2021).

Deliberative chatbots have been found to increase participation and lead to finding more ideas and issues through discussion (Hadfi et al. 2021).