Policy

National Assembly Seminar on “AI and Digital Inclusion” Concludes... AI Services Must Reflect Accessibility From the Design Stage

TECHWORLD ·

National Assembly Seminar on “AI and Digital Inclusion” Concludes... AI Services Must Reflect Accessibility From the Design Stage
Participants in the National Assembly seminar on “AI and Digital Inclusion” pose for a commemorative photo. [Photo: Kim Seung-gi reporter]

✦ AI Summary

The National Assembly seminar on “AI and Digital Inclusion” was held on the 10th at the National Assembly Members' Office Building in Yeouido, Seoul.

Ahead of the government’s official launch of “AI for All,” the seminar discussed procedures for people with disabilities to directly test services and for problems to be corrected after major functions are checked with assistive devices.

It also featured accessibility technology demonstrations by EquiforAll and Meta Korea, along with proposals to improve accessibility in public infrastructure, complaint systems, kiosks, and generative AI services.

BizCrushWant to follow Korean talks and live streams in real time?Try CapCrush on the web

The National Assembly seminar on “AI and Digital Inclusion” concluded on the 10th at the National Assembly Members' Office Building in Yeouido, Seoul. At the seminar, held ahead of the government’s official launch of “AI for All,” speakers called for accessibility checks that reflect the actual usage environment of people with disabilities. Lawmakers proposed a process in which people with disabilities themselves directly test services, and the discussion continued with the view that major AI service functions should be tried using assistive devices such as screen readers, with problems found in the process then corrected.

The seminar opened with accessibility technology demonstrations by EquiforAll and Meta Korea. EquiforAll demonstrated an AI avatar that delivers text information in Korean Sign Language. It also introduced a case in which a Jeju disaster text alert link was used to check sign-language disaster guidance, as well as a case in which food-packaging “Food QR” was used to check consumption precautions in sign language.

Meta Korea demonstrated AI glasses that recognize the surrounding environment and objects with a camera before describing them by voice. The AI glasses worked by taking the user’s question and then providing a voice explanation of what it had recognized. The event highlighted concrete use cases in which AI helps people with disabilities access information, such as sign-language delivery of text, disaster, and food information, and voice descriptions of the surrounding environment.

The seminar was hosted by Rep. Seo Mi-hwa of the Democratic Party and jointly organized by the Korea Federation of the Blind and the Codit Global Policy Research Institute. Held after the Digital Inclusion Act took effect in January, the seminar served as a forum to discuss policies needed for the development and dissemination of accessibility technologies.

The keynote presentation was delivered by Prof. Lee Hyun-gyu, director of the AI and SW Education Center and professor in the Department of Computer Science and Engineering at Incheon National University. He stressed the need to examine not only AI performance but also where it is used and who benefits from it.

He called this direction “AI for Good.” As principles for “AI for Good,” he presented five values: public interest, accountability, fairness, safety, and inclusiveness.

He also explained that the social positive effects of technology need to be reviewed as well. He said the core of digital inclusion lies in guaranteeing the enjoyment of basic rights.

The direction presented groups the conditions that make digital use possible into three pillars: accessibility, digital capability, and the use of opportunities. For this, he said a telecommunications network and device usage environment are needed, as well as the ability to identify fake information and phishing. Only when those conditions are in place, he explained, can the use of digital technology lead to employment, income-generating activities, and participation in public administration.

As measures to expand public infrastructure, he proposed broader public Wi-Fi access and improvements to the environment for using assistive devices for people with disabilities. As the role of digital learning centers and experience centers, he suggested training on how to use devices according to different types of disability. He also included trial buses as a support measure for residents in remote or mountain areas and for people with disabilities who have difficulty traveling.

He also proposed real-time sign-language interpretation, braille output, and voice guidance as standard features in public complaint systems. He suggested improving complex administrative forms by writing them in plain language, and also raised the need to support people who have difficulty handling online civil petitions. He further proposed placing sign-language interpreters and digital guides at community centers and public institutions.

He shared an experience of using a kiosk and then going to a store counter because the screen operation sequence was different from what he expected. He pointed out that some users cannot reach the screen or need voice guidance, and that if face-to-face counters disappear, those users may have difficulty placing orders. He said he had personally failed to place an order at a fast-food restaurant because the kiosk was difficult to use.

He said accessibility needs to be considered from the service design and planning stages. As ways to reduce this problem, he proposed introducing tactile keypads or physical buttons, and also suggested using voice prompts to guide the next step when an operational error occurs. In this way, he explained, even people who have difficulty operating screens because of presbyopia, shaky hands, or temporary injuries can use the service, and companies can also expand their customer base.

He also proposed ways to use generative AI. He mentioned a method in which AI recognizes a medicine bottle or a sign and then provides guidance by voice or sign language, and another in which people can ask in everyday language about the date of benefit payments without going through complicated menu procedures.

He also suggested that, for such services to function in practice, the training data should reflect the characteristics of vulnerable groups such as people with disabilities who have traditionally been underserved by services. He also proposed expanding the acquisition and opening of public data related to accessibility.

The earlier proposals divide the roles of the government, companies, and civil society in securing accessibility. He proposed that the government establish accessibility guidelines and provide tax support. He also urged companies to design with accessibility in mind from the planning stage. For civil society, he suggested a role in checking whether services actually work in the field.

Following these discussions, the plenary session focused on accessibility issues in generative AI services. The session was moderated by Min Kyung-ki, visiting professor at KAIST Kim Jae-chul Graduate School of AI.

Han Hye-kyung, a researcher at the Korea Federation of the Blind, said she was a person with a visual impairment herself. She explained that screen readers read all multiple operation items only as “buttons.” She also said users have to memorize the order of the photo attachment button and the message send button.

She pointed out that the guidebook for the “AI for All” project contest includes accessibility improvement measures as an evaluation item. However, she noted that it does not explicitly require a process in which people with disabilities directly use the service before the official launch.

She argued that digital services are the same as installing an elevator after a building has been completed, since space and traffic flow then have to be secured again. She stressed that AI services should be systematically designed from the earliest planning stage to accommodate different disability types. She also emphasized the need to conduct user evaluations before launch.

Lee In-gu, CEO of EquiforAll, explained that expert involvement is greatly needed both for sign-language translation and for building training data. He said draft sign-language translations generated by LLMs require expert review, and that experts must also revise sign-language avatar movements. He added that many experts are involved in building sign-language training data, and that the process requires individual tagging by sign-language experts who review the video content.

In line with this, he proposed integrated support linking R&D, data building, verification, and market development. He said equal access to information is the beginning of a fair-opportunity society.

Heo Wook, vice president of Meta Korea, said the company is working with Naver Map by reflecting feedback from visually impaired people in Korea who tried the AI glasses. He said trial feedback included a request for voice navigation guidance. He then stressed that participation, evaluation, and opinions from the people concerned are the most important factors.

It was noted that existing web accessibility standards can be used to check the formal elements of chatbots, but are insufficient for addressing the variability of generative responses and the quality of interaction. Kim Na-jeong, a legislative researcher in the Science, Broadcasting and Communications Division of the National Assembly Research Service, said existing web accessibility guidelines can be used to check chatbot input fields, but that the existing guidelines have limits when it comes to evaluating responses and interaction quality, which vary by user.

As a complementary measure, Kim Na-jeong, a legislative researcher in the Science, Broadcasting and Communications Division of the National Assembly Research Service, proposed using the Digital Inclusion Act’s basic plan and impact assessment to coordinate accessibility policies across ministries. She also proposed making it mandatory to reflect the opinions of people with disabilities in impact assessments.

Hong Kyung-sun, senior researcher of the Digital Accessibility Team at the National Information Society Agency, pointed out that if AI gives incorrect medication information and users have difficulty checking the response, it could lead to misuse. She added that the agency is preparing accessibility guidelines to apply from model development through service operation and use, and that the goal is to establish an organizational standard by the second half of next year.

The discussions on the day led to procedures for verifying the actual use environment of new technologies such as sign-language avatars and AI glasses. In the “AI for All” project, tasks remain as to whether people concerned will be able to review the accessibility improvement plans presented at the proposal stage before launch, and whether problems discovered after launch can also be corrected.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=409533

References

This article was produced with the help of an automated content generation algorithm.


Source: TECHWORLD

View original

This article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.

Keep reading

BizCrush
BizCrush showing live captions over an online meeting

BizCrush app

Live captions and summaries for meetings and calls

BizCrush Inc. · CEO Taemin Kwak

HQ 100 Middlefield Rd, Menlo Park, CA 94025, USA

Korea office Z112, Room 602, 6F, 602 Yeongdong-daero, Gangnam-gu, Seoul, Republic of Korea

Support help@bizcrush.ai · Partnerships business@bizcrush.ai

AI Notice

Some features use generative AI.

Those features carry their own pre-use notice and mark AI-generated results. AI output is for reference only — always check the original source and consult an expert before making an important decision.