Ethics Framework in AI and Data Utilization - Introduction material -

Japanese Version
https://speakerdeck.com/jdmc/aidetahuo-yong-niokerulun-li-huremuwaku-goshao-jie-zi-liao

Transcript

1. Copyright 2023 Japan Data Management Consortium 1
   April 2023.
   Ethics Framework
   in AI and Data Utilization
   - Introduction material -
   Compliance Study Group for AI and Data Utilization
   

   View Slide

2. Copyright 2023 Japan Data Management Consortium 2
   Ethics Framework in AI and Data Utilization
   What are stumbling points What is Ethics Framework
   Recently, there has been an increase in
   efforts to utilize data acquired through
   business to provide new value to
   people and society.
   However, in data utilization, actions
   taken with good intentions can often
   become problematic.
   Even if there are no legal issues,
   situations that cause public backlash
   may happen.
   These problematic factors are referred
   to as "stumbling points".
   When attempting new things through AI and data utilization, it is
   important to find stumbling points in advance and take measures to
   address them.
   However, simply confirming based on legal regulations is insufficient,
   and it is necessary to find them from an ethical standpoint.
   The "Ethics Framework" is the most appropriate content in modern
   times, which analyzes recent problem cases and organizes them into
   an easy-to-understand system.
   The Ethics Framework is organized
   into five ethical items
   in the figure on the right.
   Below the ethical items are
   14 stumbling points and
   49 checklist items.
   1 Propriety
   2 Accountability 3 Fairness
   4 Safety and security
   5 Information protection
   

   View Slide

3. Copyright 2023 Japan Data Management Consortium 3
   Kakogawa City’s Advanced SAFETY CAMERAS
   - Pre-AI Operational Verification Based on Ethical Framework -
   Efforts of Kakogawa City
   Since FY 2017, approximately 1,500 SAFETY CAMERAS
   have been installed.
   Installed mainly in the school commuting routes,
   school surroundings, and park surroundings of elementary schools.
   Considering residents' privacy.
   Strengthening communities through local monitoring of children and the elderly.
   Achievements so far
   Crime prevention effect and early resolution of crimes.
   The number of crimes in the city has been less than half
   in about four years since the installation.
   In FY 2022, 150 AI-enabled Advanced SAFETY CAMERAS
   were installed.
   The aim is to further create a safe and secure town.
   Detecting unusual sounds such as screams,
   and warning the surrounding area using a speaker and a Rotating light.
   Detecting dangerous driving behavior of automobiles
   and alerting pedestrians through warning lights.
   Automatically grasping traffic and pedestrian flow data
   and utilizing it for city planning considerations.
   Verification through Ethics Framework.
   Verification results.
   By using the 14 "stumbling points" and 49 check items,
   we can comprehensively check the response situation
   from an ethical perspective.
   Confirmed that appropriate measures have been taken
   for all check items.
   Flow from data acquisition to utilization
   Camera body
   Intermediate server Data linkage platform
   Acquired data New value
   Image / Audio
   Unusual sound
   Vehicle approach
   pedestrian traffic
   Town planning
   considerations
   Image / Audio
   Used for AI training
   for unusual
   sound detection.
   Crime prevention
   (Rotating lights and speakers)
   Accident prevention
   (Rotating lights and speakers)
   Unusual sound
   Vehicle approach
   pedestrian traffic
   Non-processed data
   *To be discarded
   Processed data
   ※No personal
   information included
   Derived data
   

   View Slide

4. Copyright 2023 Japan Data Management Consortium 4
   Ethics
   1 Propriety
   1.1 Satisfaction with data utilization
   1.2 Easing concerns with providing personal information
   1.3 Balance between the risks and benefits of providing personal information
   2 Accountability
   2.1 Explanation of derived data results
   2.2 Potential for biases and errors in derived data
   3 Fairness
   3.1 Possibility of feelings of unfairness and discrimination based on derived data
   4 Safety and security
   4.1 Potential impact of derived data on individuals, organizations, and local communities
   4.2 Potential for abuse
   4.3 Data stored/managed in Japan
   5 Information protection
   5.1 Information security measures
   5.2 Control at provision/publishing point
   5.3 Appropriate deletion of collected data
   5.4 Identification of individuals by matching with other data
   5.5 Infringement of rights relating to data provided
   stumbling points
   Ethics
   stumbling
   points
   1 Propriety
   1.1 Satisfaction with data utilization
   1.1.1 [Propriety of purpose]
   Is the purpose of data utilization generally
   appropriate?
   1.1.2 [Propriety of collection]
   Is the data collected for the purpose appropriate
   in general?
   1.1.3 [Propriety of generation method]
   Is the method for generating derived data from
   collected data generally considered appropriate?
   1.1.4 [Propriety of classification or meaning]
   Are the types and implications of the generated
   derived data generally considered appropriate?
   1.1.5 [Propriety of how data should be used]
   Is the use of derived data generally considered
   appropriate?
   1.1.6 [Explanation of propriety]
   Are you explaining to data providers so that they
   understand the propriety of these practices?
   checklist items
   The Ethics Framework
   14 points
   49 items
   

   View Slide

5. Copyright 2023 Japan Data Management Consortium 5
   

   View Slide