JST 22:00~24:30 / CET 14:00~16:30 / UDT 13:00~15:30 / EST 08:00~10:30, January 4, 2022
A personal AI (PAI) belonging to each of us could maximize the value of our personal data by fully utilizing the data to enhance the quality of our life and work, while disclosing the data to others in limited cases. The whole value created thereof will be enormous. The production and operation of such PAIs therefore would be by far the most profitable business. This PAI business is likely to prosper within a decade, aggregating our personal data to each of us to let our PAIs utilize our data to assist or perhaps control us. Most of our actions of selecting and purchasing something would then follow our PAIs’ decisions because we are all lazy without exception. Governance of PAIs is hence an essential issue concerning our privacy, human rights, welfare, economy, politics, culture, and so forth.
This symposium will shed light on benefits and risks of PAI from diverse viewpoints encompassing computer science, marketing science, ethics, legal science, sociology, among others. Below are some topics to be addressed.
- technical foundation of PAI
- social acceptability of PAI and PAI business
- governance of personal data using PAI
- governance of PAI for both personal and societal values
- role of PAI in Society 5.0
The symposium consists of the following two parts, each of which comprises some presentations followed by discussions involving the audience.
Part 1: Personal-Data Ecosystem and Its Social Acceptability
Koiti Hasida: Personal-Data Ecosystem
Suppose you have your own personal AI (PAI) exclusively dedicated to you. It exclusively utilizes all your personal data (PD) to intervene in your life more deeply and carefully than any other technologies.
When B2C servicers S1 … Sn (such as Google, Amazon, a government, a school, a supermarket, a coffee shop, etc.) use your PAI as the digital customer-contact point, the PAI can incorporate the functionalities of CAI1 … CAIn (the centralized AIs operated by S1 … Sn) among others. Also it can access all your PD available to S1 … Sn as well as other PD generated by other services and sensors. Using more functionalities and more PD than the servicers, the PAI provides much richer services than S1 … Sn in total.
For instance, the PAI can use CAI1 and the PD available to Sn to provide a service which neither S1 nor Sn can provide. This may be a healthcare service using exercise data, a purchase-assistance service using health data, and so on. Note that the PAI does not disclose this PD to S1. Neither is the technology of CAI1 disclosed to Sn. The profit from this service will be shared among the stakeholders including S1 and Sn, so that the customers and the servicers all benefit from the PAI: a value cocreation by all the stakeholders.
Keiko Toya: A Study of Consumer Acceptability for Personal Data Service
Digitalization has been accelerating in the COVID-19 pandemic. Services so far offered face-to-face by people only are becoming possible to offer online and/or automatically due to the advance of technologies such as IoT and AI. The proper use of customers’ data will make these services more valuable for both customers and companies.
The study aims to clarify the structure of consumer acceptance of their data usage. By a survey in the medical and financial fields, we verify two hypotheses. One is that knowledge and involvement affect consumers’ information processing strategy according to the ELM (Elaboration Likelihood Model). We divide the consumers into segments by the level of knowledge and involvement. The other hypothesis concerns what consumers prioritize for their acceptance of their data.
Based on a qualitative analysis, we focused in our survey on three factors: 1. Smartphone-internal computation. 2. The improvement of service by using personal data. 3. The trustworthiness of the service providers. Our survey measured the difference in acceptance of data usage among participant segments using the ELM. We collected the data and are analyzing it now. We will present the result at the symposium.
Part 2: Potentials and Governance of PAI
Koiti Hasida: From Centralized AI to Personal AI
Since global nuclear war and global warming have no winners and inflict massive damage on all humanity, it would be possible for the entire world to work together to avoid them. On the other hand, there is a widely recognized risk that technologies such as AI and biosensing may promote surveillance capitalism and totalitarianism, threaten human dignity, and damage democracy, but it will be difficult for humanity to jointly manage this risk because there will be winners in the race to develop these technologies.
As PAI creates much greater value than CAI, servicers will use PAIs instead of CAIs as digital customer contacts for the sake of larger profits, which will dissolve the above problem of how to restrict CAIs. Many servicers’ employment of PAIs entails the spread of decentralized management of PD. That will allow governments, universities, research institutes, NPOs, private firms, etc. to collect PD directly from many people and analyze it to monitor PAIs, also monitoring each other to raise and maintain their trust. Thus emerges a governance of PAIs by trusted parties, to implement a value cocreation in a proper balance among individuals, service businesses, and local & global societies. This will raise the social acceptability of PAIs, letting more servicers shift from CAIs to PAIs. Authoritarian governments may keep using CAIs, but most servicers (both public and private) around the world will employ PAIs. CAI-based authoritarian regimes will no longer spread.
Ayako Kato: Use-Cases of Data Integration by Individuals
1) As a use-case, we can say that individual side’s data aggregation is useful for realizing Smart-city or MaaS. A concept of ‘City-OS’ as a data linkage platform in each city or area has been suggested, however, this is a centralized system. The limited cities and service providers which are engaged in such a certain system can be connected. On the other hand, if we aggregate and utilize our own personal data for ourselves using Personal AI (PAI), then our transports may become smarter.
2) Decentralized data utilization with PAI needs an intermediary which mediates between individual side and service provider sides. Thus, an intermediary in a decentralized system has two-sided markets. Who can develop such intermediaries and PAI? IT-platform business operators such as Google, Amazon, Facebook, Apple (GAFA) and the like, which have numerous accounts can be actually intermediaries for decentralized personal data utilizations and developers of personal AI (PAI) agents.
3) Even if personal data is handled by each individual using his/her PAI, intermediaries may obtain transaction data regarding each personal data usage. Decentralized personal data utilization with PAI might be able to lead to centralized data integration into a few IT-platformers. Therefore, it might be necessary to make rules for individuals to clarify what kind of rights they have. This presentation will indicate some issues for the personal data ecosystem.
Hiroshi Nakagawa: An Architecture of PAI Agent — Four types of trust
The personal AI (PAI) agent is an AI program that reduces the cognitive load on the data subject by intervening between the individual data subject and the service provider in the external information environment via the Internet. In order for this intervention to be effective, the PAI Agent and the external service provider must trust each other. Furthermore, of course there is a need for a mechanism that allows mutual trust between the PAI Agent and the data subject who uses the PAI Agent . These two types of mutual trust relationships must be established automatically. In other words, the individual data subject only has to make a very small amount of effort. Many of the trust relationships have become feasible because many of the mechanisms for establishing each trust relationship and the work of international standardization have already been proposed and implemented. In this presentation, we will give an overview of how to establish a trust relationship that has already been worked out by many IT researchers and developers, and explain the issues that remain for the establishment of a full trust relationship.
Takafumi Ochiai: Governance of PAI and Governance Innovation in Society 5.0
Ministry of Economy, Trade and Industry “GOVERNANCE INNOVATION Ver. 2: A Guide to Designing and Implementing Agile Governance” (pⅲ)
The Japanese government is aiming to achieve a human-centered society where high-level integration of cyberspace and physical space can promote economic development and solve societal issues. Implementing innovative technologies including management of personal information in society will require a fundamental reform of the governance model, in light of changes in social structures brought about by new technologies and frameworks.
Governance for the personal AI (PAI) agent will also prepare its appropriate governance system which is based on not only government regulations but also self-regulation system to protect individual rights and profits for related parties providing services to individuals. I will introduce discussions of forms of “agile governance” that are designed to continuously and rapidly run cycles of “conditions and risks analysis,” “goal setting,” “system design,” “operations,” “evaluation,” and “improvements” with multiple stakeholders in various governance mechanisms, such as corporate governance, regulations, infrastructures, markets, and social norms in order to provide basic consideration for discussion about future governance system of PAI.
Organizers
- Koiti Hasida (The University of Tokyo & RIKEN)
- Keiko Toya (Meiji University)
- Ayako Kato (Toyo University)
- Hiroshi Nakagawa (RIKEN)