Skip to main content
Recently, we are beginning to see new possibilities and issues in a new style of democracy where “citizens think for themselves and participate in policy decisions.” At Hitachi, we formed a hypothesis that “Each citizen can improve their own well-being through taking the initiative to participate in the political process and gain a sense that they can determine their own future,” and are engaged in developing AI technologies that support cyberdemocracy, such as balanced acquisition of information and simulation of policy outcomes to support citizens’ participation. In this dialogue, Professor Masaki Taniguchi of the Graduate Schools of Law and Politics at the University of Tokyo, joins Toshihiro Kujirai from the Research & Development Group of Hitachi, Ltd., to discuss citizens’ participation in policy-making looking at the “current situation,” “the ideal state to be attained,” and “issues,” from their respective positions as a political scientist and AI researcher.
(Published 26 August 2022)
We are shifting from a one-way delivery of information by governments and corporations to citizens, to a bi-directional exchange of information where citizens may be the ones providing information. What sort of issues are there for citizens to receive correct or fact-based information from among the massive volume of information available at large?
Let’s use food as an example for information, and imagine eating lunch every day at a company or university cafeteria. In the past, the company cafeteria had a more limited menu than now, for example “Western-style meal or Japanese-style meal,” or “set menu A or B,” with just one or two choices in the daily menu, which is what many people ate. While the menu choices were limited, the set menus had a main dish as well as side dishes, or the main dish alternated between meat and fish depending on the day of the week; and the preparation could be simmered, grilled, or fried. In the long term a nutritionally balanced menu was provided.
In the world of communication about politics, we can equate newspapers and television news to these set menus. These media supply information about sports and entertainment, but also include significant portions of hard news on politics and economics. In such a way, the reader or viewer is exposed to serious topics.
Times have changed, and the company cafeteria now offers a rich selection of menus. This is probably the same at Hitachi. Many places allow each person to choose the kind of dish they like in a cafeteria-style setup, don’t they? This means that a person who loves meat can eat only meat today, tomorrow, and the day after. Or someone who likes noodles can choose ramen today, udon wheat noodles tomorrow, and soba buckwheat noodles the next day—never eating fish or vegetables. Such choices are now possible. While each person’s preference may be satisfied, unless one is careful, it may result in nutritional imbalance.
Something like this is occurring today in the world of political communication. The number of television channels has grown exponentially with advances in satellite broadcasting and cable television. News flows out from terrestrial television just by leaving the set on. But with specialized channels on satellite broadcasting or cable television, viewers can pursue only their interests, such as films or sports, all day long.
The popularization of the internet has exaggerated this tendency. Conservatives in America receive their information from newspapers, television, and radio that espouse the same way of thinking, and they exchange opinions on internet sites that draw conservatives. Similarly, liberals fill their surroundings with information from liberal views. We call this “selective exposure” whose consequence is an “echo chamber effect.” The result of taking in only information that sides with one’s point of view has hardened the ideology of conservatives and liberals even more, sharpening political polarization and confrontation.
One of the factors accelerating the echo chamber effect could be the “filter bubble.” Algorithms on internet search sites block information that is not of interest to the user and allow only information that the user wants to see. On one hand this creates a comfort zone for people with similar opinions, but on the other hand it presents the disadvantage of losing exposure to different viewpoints, similar to what you mention is happening in the political sphere. Are there other issues?
A serious issue is “fake news.” What is distributed is not only incorrect information but false information that has been created from the beginning for the purpose of deceiving people.
There is a lot of fake information that is circulated to fan political divisions. In the Russian invasion of Ukraine, there have been deep fake videos circulated that one could mistake for being real.
Fact checks are conducted to differentiate what is true and what is not, but checking each news item requires much time and is costly. Moreover, even if a journalist examines what a certain politician said and determines that it is fake news, the supporters of that politician won’t believe that it is fake. There is research that shows that those people choose to believe what the politician said was true and “it’s journalism that is fake.”
The non-profit organization, Open AI, founded by Elon Musk and others as an artificial intelligence research laboratory, released in 2020 “GPT-3” which can automatically generate sentences, and in 2021 “DALL-E” which can create images from textual descriptions. While the use of such technological advances to automatically generate text in email and articles and to write computer code has progressed, there is concern that massive amounts of incorrect information could be disseminated over the internet.
When we shop on the internet, our search and purchase history is filtered, and we might see recommendations for products that fit our preferences. This is convenient for shopping, but it can result in an unfortunate linkage for democracy and technology.
In a democracy, what is important is the process of listening to other opinions without making quick conclusions, engaging in discussions, and reflecting on them. AI and other new technologies can be effective if they can organize varying opinions and deliver these to people in an easily understandable way, and offer proof of the accuracy of the underlying information.
AI is skilled at finding needed information from among a flood of information and organizing information according to different points of view. It can support the disseminating and receiving of information.
At Hitachi we have developed technology that can specify with a high degree of precision the reasoning for and background information of summarized statements that exist on the internet by searching for corresponding original documents.*1 We think this allows the user to trace the original source of the information and more easily assess the reliability of the information.
What other areas are being researched?
We are conducting research and development on a technology that will organize and show texts from multiple perspectives, such as economics and public safety, regarding a particular policy by selecting them from a large volume of data sources.*2 We have high expectations that this type of technology will allow gathering of information from multiple perspectives, without being biased toward any particular point of view.
As information on society and politics becomes more prevalent, what are your thoughts on how we can extract the real opinions of citizens?
What first comes to mind is the public opinion survey. Even when we believe it was conducted in a fair and neutral manner, there is an implicit bias. For example, we can imagine that “the result of asking 100 white collar workers at Shinbashi Station” or “the result of asking 100 young people at Shibuya’s Hachiko statue” does not reflect the public opinion of all Japanese. In recent years, the response rate of even governmental public opinion surveys is declining, leading to a serious effect of a non-response bias.
There is also the problem of getting at “what is the real opinion” in public opinion surveys. People may be asked, “Are you in favor of the draft legislation on XX or are you opposed to it?” But not all of them are necessarily knowledgeable about the content of the proposed law.
Their responses may be different depending on whether they have heard the opinions of proponents of the law or those opposed to it. Naturally, perspectives differ depending on whether the law benefits the person or not. The type of opinion we want to find out is not the aggregate of the instantaneous responses given in a telephone public opinion survey. We want to know the aggregate results of the final decision made after respondents have contemplated sufficient information.
When a decision is made after deliberation, it can lead to a sense of choosing one’s own future. Even if one’s opinion isn’t the one that is enacted, understanding the merits and demerits of the opposing opinions should make it easier to accept the result.
The American political scientist James Fishkin has come up with the “deliberative polling” method to measure public opinion resulting from deliberation. For example, after a conventional public opinion survey is conducted, participants are selected from among the respondents to attend a meeting. They are divided into small groups to deliberate materials that are handed out and given time to ask questions of specialists. After this, the change in their opinion is surveyed.
There are many hurdles that must be overcome to put deliberative polling into practice. Holding the meetings incurs high costs, and it is necessary to take precautions to ensure the meetings are conducted in a fair and neutral manner. Would it be possible to facilitate people’s deliberation and discourse by leveraging AI?
I think there are three points that must be cleared in order for people to have the sense that they are deciding on their own future through the process of deliberation and discourse.
First is whether they understand the information needed for discussion and decision. Through AI technology, it is possible to collect the relevant information from among the huge volume of information scattered on the web. It can sort and display various opinions about the proposed laws and government policies that are up for discussion. When fact-checking technology is combined with this, it will be able to show accurate information which can support the understanding of information necessary for discussion and decisions.
Second is whether they can state their own opinions after deliberation. On this point, I think the use of AI technology will allow us to show various future scenarios to study them in advance to find problem areas and possibilities for solutions. Hitachi is advancing the use of “Foresight,”*3 a discussion support tool for depicting future scenarios. It is also pursuing the use of “AI that supports policy proposals,”*4 a tool to examine the policies necessary to realize future scenarios. If we can depict diverse scenarios and have an image of what kind of future we desire, it will deepen our understanding of the draft legislations and policies that are the subjects of discussion. It can also promote the understanding of necessary information. We should also be able to see the effect of people voicing their own opinions arrived at after deliberation.
And third is, whether it is possible to confirm that the opinion is reflected in some form. I think we can take into account various opinions and arrive at a consensus if we can facilitate discussion in a neutral manner without losing diverse perspectives. We can support facilitation and automation by using AI in monitoring the perspectives under discussion and the speakers on those matters. Further, by continuous monitoring and informing citizens of the effects of the policy decided upon, they will be able to find out about the result of their selections.
How should the digitalization of politics proceed in order for citizens to be active participants in policy decisions?
Article 56 of the Japanese Constitution states: “Business cannot be transacted in either House unless one-third or more of the total membership is present.” As to the interpretation of “present,” the Commission on the Constitution of the House of Representatives has made the interpretation that “in case of an emergency situation, an exception will be made to allow being present online.” In order to take the next step, heightening trust in security is needed for identity verification of legislators to prevent identity fraud, and to defend against cyberattacks.
In Japan, public agencies use the “my number” card to verify identity and a password for authentication. But, with this authentication method, there is the risk of safe-keeping the card and unauthorized reuse of the password. At Hitachi, we are pursuing research and development of finger vein authentication*5 which is considered to be a highly precise and convenient method of biometrics authentication. We provide a safe and easy-to-use Biometric Signature Server as part of the Public Biometric Infrastructure (PBI).*6
At the root of deliberative polling and digital government is the concept of deliberative democracy. Jürgen Habermas, a leading theoretician, declares that in addition to the official political round of representative democracy deciding on policies after representatives hold discussions in the national parliament, a secondary round of non-official deliberations in the public sphere should be activated for effective policy decisions in democratic politics. His thought is that this would raise new problems and solutions that were not attended to in the national parliament, and that a thorough discourse on opinions would lead to better acceptance of the decisions, bringing democratic politics closer to the people and making it stronger.
What kinds of technologies are needed for putting into practice a deliberative democratic political process?
First, can’t we develop tools to support the aggregation of opinions when thoughts about a certain measure are surveyed?
Second, solidify the justifications for requiring legislation by analyzing big data scattered around in society and present forecasts for the ways society will change under policy initiatives. If deliberations by the people and the legislative bodies can be advanced by providing information that includes verification of the need for legislation stated by the government, that would be wonderful.
And third, who are the legislators who might be most likely to listen to one’s opinion, even if it isn’t the legislator representing the local area? Have there been discussions in the past on the issues one is concerned about? I would like a function that would perform a cross-sectional search and matching of meeting minutes, records of statements, and public pledges by legislators. As some legislators don’t have opportunities to make statements, there is a limit to searching parliamentary meeting minutes; and the data archives of mass media are dispersed among the companies, restricting the portions that can be searched. While covering all the information, disinformation and slander need to be removed, making this a process that is “easy to say, but difficult to do.”
I hope that this AI technology’s support for decision-making by legislators and citizens as well as its promotion of interaction between them, can show society that technology can work for the progress of democracy.
I would like to consider the possibilities of support by AI for the three proposals you have given us.
The first is support for aggregating opinions about policy measures. For the extraction of opinions about specific policies from social media and blogs, and gathering proactive opinions, we might be able to utilize telephone surveys using chat bots and voice recognition.
The second is support by AI for policy proposals. When we simulate various future scenarios, we know that there will be changes depending on which policies are implemented. I think that by focusing on that junction point AI can support the verification of legislative facts.
The third is support in searching for politicians who are supportive of one’s viewpoint by going through various documents. It will be possible to find legislators whose opinions are close to one’s own by searching through the records of statements they make in meetings, public pledges, and websites. It will also be possible to find legislators who state differing opinions, and compare them.
AI can generalize past experiences, but it is difficult for AI to describe new situations. As we advance research and development to solve these technological issues, we hope to create a way that people and AI can collaborate by complementing each other’s strong and weak points to build a new future that cannot be depicted by either on their own.
I appreciate being able to talk to you about the possibilities of changes in the way politics is carried out using technologies that include AI. While we can see the beginnings of the implementation of those changes, I think we were able to discuss what needs to be done in terms of advancing our research. That said, in order to realize this new type of politics, it is not only technological development that is needed. I see the need for collaboration among political science researchers such as yourself, those who enact and implement the policy proposals in national and local governments, as well as companies like Hitachi that provide technology.
Our research and development efforts work toward collaboration between people and AI, and support for cyberdemocracy, which is a technology that can find what is accurate from among a large volume of information; a technology that can organize and give an overview of information from various perspectives; a technology that can simulate the future; and a technology that can facilitate discourse from a neutral position. As we collaborate with others, we wish to contribute toward bringing about a society where its members can decide on their own their image of the future and make their own selections.
When I sought the opinion of University of Tokyo Professor Emeritus Jun’ichi Tsutsui, he said, “AI in Japan is not for surveillance of the population by the government. Neither is it solely for corporations to increase their profits. AI in Japan can become a partner in cooperating with people, assisting people, and looking out for people.” This rang true for me. As we are living in the country that created the robot characters Atom Boy and Doraemon, we can develop a Japanese form of “AI that works together with people.” To draw it towards my field, I hope “AI that advances democracy” can be developed.
Just as digitalization (DX) is increasingly implemented in the industrial world, it is also reaching the political sphere. In Japan as well, discussions are being held on a political system that will suit our present information society, with expectations for leveraging AI in that effort. Hitachi’s AI technology will contribute to the realization of a society in which we citizens can participate more actively in the political process, and contribute to a vision for the future of their society.
(As at the time of publication)
TANIGUCHI Masaki, Ph.D.
Professor, Graduate Schools for Law and Politics, The University of Tokyo
President, Nippon Institute for Research Advancement
Masaki Taniguchi is a professor of political science at the University of Tokyo and a leading scholar in contemporary Japanese politics, electoral studies, politics and mass media.
He has authored numerous books including, Representative Democracy in Japan (University of Tokyo Press), Politics and Mass Media (University of Tokyo Press), and The Theory of Party Support (Iwanami Shoten). His publications in English appear in journals such as, the International Political Science Review, Political Communication, Electoral Studies, Journal of Elections, Public Opinion and Parties, and Japanese Journal of Political Science.
Masaki also serves as the president of the Nippon Institute for Research Advancement (NIRA), one of the most prestigious think tanks in Japan, and commissioner of the National Commission for the Management of Political Funds nominated by the Japanese Diet. He received his B.A. and Ph.D. in political science from the University of Tokyo.
KUJIRAI Toshihiro, Ph.D.
Head of the Media Intelligent Processing Department
Advanced Artificial Intelligence Innovation Center
Research & Development Group, Hitachi, Ltd.
Toshihiro Kujirai joined the Central Research Laboratory of Hitachi, Ltd. In 1997 after completing his Master of Aeronautics and Astronautics. As a researcher, he pursued R&D for digital technologies in areas such speech recognition, HMI, remote sensing, data mining, and reinforcement learning. Currently, as the head of research for media intelligent processing, he leads research activities on NLP, knowledge processing, speech recognition, acoustic diagnosis, explainable AI (XAI), and so on.
Toshihiro received his doctoral degree in engineering from Tottori University. He is a member of The Japanese Society of Artificial Intelligence and a member of SC42 at Information Technology Standards Commission of Japan in Information Processing Society of Japan.