Associate Professor Daniel Berliner from the Department of Government at the London School of Economics and Political Science (LSE) delivered a lecture titled "Information Processing in Participatory Governance" on April 25th. This lecture, part of the Democracy and Democratisation in East and South Asia course, was jointly organized by the Graduate Institute of Development Studies (GIDS), the College of Social Sciences’ Internationalization Project, and the 50th anniversary lecture series of GIDS. The event was hosted by Prof. Mei-Chuan Wei.
Berliner’s lecture focused on the crucial role of participatory governance in democratic practice, particularly on information processing issues within participatory governance. His presentation was structured around public participation in decision-making processes, the importance of information processing, a typology of information processing, issues of costliness and delegation, conclusions, and future research directions. Berliner emphasized that his research is based on empirical data and aims to establish typologies of information processing models in public policy decision-making.
Regarding public participation in decision-making, Berliner explained that it involves the general public's involvement in policy formulation and implementation through various forms such as town hall meetings, consultation meetings, petitions, participatory budgeting, deliberative mini-publics, civic reporting, and complaint hotlines or platforms. His research and lecture focused on processing information that helps decision-makers understand public preferences, perspectives, issues, and potential solutions.
Berliner highlighted that existing literature, such as Fung’s research, explores public participation from aspects like inclusiveness of participation, deliberativeness of interactions, and linkage to public authority. Other studies analyze participation from empowerment, institutionalization, scope, and spatial scale (central or local) perspectives. He pointed out that common criticisms of public participation include lack of representativeness, accountability, and mere window dressing. His research specifically examines situations where decision-makers or participants genuinely care about learning from useful information, regardless of whether this concern arises from policy considerations or political motives.
To illustrate the difficulties of public consultation in government decision-making, Berliner cited the example of UK Deputy Prime Minister Nick Clegg's attempt to involve the public in legislative reform through an online platform. However, the initiative had to be abandoned due to the overwhelming number of 46,000 responses, making information processing an impossible task. Berliner stressed the importance of effective information processing methods, as public participation is crucial for government policy responsiveness.
Berliner’s analytical framework focused on information inputs, information processing, and information outputs. He emphasized that not all information beneficial for government decision-making is aggregative. His typology based on the dimensions of specificity and novelty categorizes information into four types: aggregative, classificative, summative, and generative. He explained that specificity differentiates information produced by aggregating opinions or filtering, while novelty determines whether the produced information is previously known or unknown.
Discussing the concept of generative information, Berliner noted that this concept was adopted before the widespread attention to ChatGPT. He further explained that aggregative information is low in both specificity and novelty, summative is high in novelty but low in specificity, classificative is high in specificity but low in novelty, and generative is high in both specificity and novelty. Berliner emphasized that these types are constructed purely for analytical purposes and have no inherent superiority.
The lecture concluded with a focus on generative information processing. Berliner used examples such as the amendment of U.S. laws and Uganda’s U-Report system to illustrate that the produced information (including issues, perspectives, and solutions) was previously unconsidered or unknown. However, the generation of such information relies heavily on the direct and in-depth participation of relevant personnel, including supervisors, and requires domain knowledge and careful consideration from information processors, leading to higher time and effort costs. He stated that the cost of information processing is closely related to the degree of delegation. If processed by experts, elites, or participatory/deliberative groups, it would be costlier than decentralized voting or third-party methods. Berliner also discussed the potential scenarios of information processing by artificial intelligence, explaining the possibilities of natural language processing (NLP), supervised machine learning (supervised ML), and unsupervised machine learning (unsupervised ML).
During the Q&A session, attending faculty and students raised questions primarily regarding the costs involved in information processing, and inquired about specific real-world decision-making scenarios that would particularly rely on generative information. They also cited the current widely discussed issues in Taiwan, such as nuclear energy and the death penalty, to seek Berliner’s views
