Advertisement
Technology
OpinionLetters

Letters | Why Hong Kong complaints chatbot needs an upgrade

Reading Time:1 minute
Why you can trust SCMP
0
The 1823 service was launched in 2011 to handle inquiries and complaints from the public on behalf of 23 government departments. The chatbot using AI was introduced in late 2019. Photo: K.Y. Cheng
Letters
Referring to the letter from Yiwen Wang and Rui Min, we share your correspondents’ concern that most Hong Kong residents prefer to contact the government over telephone and agree that the 1823 app should be revamped to respond to inquiries in real time.

While the government has allocated HK$7.3 million (US$942,000) for developing a chatbot to serve city residents, the functions of this new tool are rather limited and services are only available in traditional Chinese.

Introduced in late 2019, the chatbot had received about 29,000 inquiries through November 2020, with a resolution rate of 90 per cent. Yet, the tool received an average rating of 3.2 out of 5 from 2,900 users, probably because the chatbot can only answer inquiries related to public transport and motoring matters.

In addition, the chatbot does not transfer inquiries it cannot handle to the 1823 staff for a real-time chat or follow-up in writing. To improve the chatbot, the government should expand the topics of inquiries this tool can handle and integrate it into the 1823 mobile phone app.

As the 1823 staff was able to handle 99 per cent of the inquiries without referring to the respective government departments, the team should be complimented for its success in knowledge management about government programmes and policies. Yet, in response to our inquiries, the Efficiency Office said the 1823 staff used Microsoft Word to keep documents on how to deal with inquiries.

03:25

How Hong Kong’s new coronavirus contact-tracing app works

How Hong Kong’s new coronavirus contact-tracing app works
We urge the Efficiency Office to abandon this 20th century technology and develop a system to manage its information in machine-readable format following the standards adopted in the government’s Annual Open Data Plans. The data can then be used to train algorithms for the automatic processing of inquiries through machine learning.
Advertisement
Select Voice
Choose your listening speed
Get through articles 2x faster
1.25x
250 WPM
Slow
Average
Fast
1.25x