Home PublicationsData Innovators 5 Q’s for Daniel Lim, Acting Deputy Director of the Data Science and Artificial Intelligence Division at the Government Technology Agency of Singapore

5 Q’s for Daniel Lim, Acting Deputy Director of the Data Science and Artificial Intelligence Division at the Government Technology Agency of Singapore

by Joshua New
Daniel Lim

The Center for Data Innovation spoke with Daniel Lim, acting deputy director of the data Science and artificial intelligence division at the Government Technology Agency of Singapore (GovTech). Lim discussed the challenges in shaping public policy with data and how GovTech is studying how other countries use open data.

Joshua New: You’ve spoken about the considerations needed to ensure data science can positively influence public policy. Could you describe some of these considerations? What seems to be the biggest challenge?

Daniel Lim: Ultimately, practicing data science in public policy means delivering analyses or software that solve real-world problems and result in real-world impact. A question that we always ask of government agencies that we consult for is: “Assuming we gave you the best piece of analysis possible, how will that change the way you design your policy or run your operations? What are the measurable improvements?” What we are looking for is work that results in practical change. We are not interested in doing research that just results in a white paper or a report.

The biggest challenge is therefore in finding these potentially “real-world impact” data science projects to work on. We put a lot of effort into working with agencies to scope data science projects—this involves: sharpening vague problem statements, such as “we wish to use AI to improve service delivery,” to specific lines of analyses with well-defined metrics; determining if data are available; and obtaining buy-in from key stakeholders who have the authority to implement real-world changes and facilitate access to the relevant data. We have an evaluation framework and only select data science projects that have high potential impact. But that is only half the battle won; we would still need to work iteratively with agencies on analyses, to tailor them for actionable insights.

All this doesn’t guarantee that every data science project will result in real-world impact, but being disciplined about evaluating the potential of proposed projects helps us define better projects and ensure that our work is relevant to the public sector.

New: Can you explain how GovTech provides data science expertise to different agencies?

Lim: GovTech deploys small teams of data scientists to agencies to collaborate on projects and support them in their own analytics transformation journeys. This also provides our GovTech-trained data scientists with the opportunity to gain exposure to applying data science across different domain areas.

New: Can you explain GovtTech’s work with the Housing and Development Board to use machine learning to analyze text?

Lim: The Housing and Development Board (HDB) has a department that receives approximately 100,000 emails each year about new flat sales, and they were interested in understanding what the key issues were. GovTech worked with this department to apply unsupervised machine learning to emails received in 2015. Through the analysis, we discovered a cluster of emails on key collection: many new flat owners were sending emails to HDB to change their appointment time for collecting the keys to their new flat. With this insight, HDB implemented an online system to schedule key collection, saving time for both the public and HDB officers in the changing of appointment times.  

Such a data-driven approach to analyzing feedback has helped improve the public sector’s “ground-sensing” ability, by surfacing emerging trends and issues that may not have been obvious before.

New: GovTech also collaborated with SingHealth, a network of healthcare providers, to try and predict hospital readmission risk. What came of this project?

Lim: This project was meant as a proof-of-concept and was completed in 2015. We had worked with SingHealth to develop a machine learning algorithm that could predict frequent admitters, who are patients admitted to hospital more than three times in a given year. The project was then taken over by the Ministry of Health, which worked with data scientists from the Integrated Health Information Systems, the healthcare sector’s IT provider. They have since developed an algorithm that would apply across all the healthcare providers in Singapore, of which SingHealth is one.

New: GovTech recently launched a study to examine open data efforts in 10 countries around the world. What are you hoping to learn from this study?

Lim: The study was conducted by the Economist Intelligence Unit and is titled “Open Government Data: Assessing Demand Around the World.” It sought to understand how citizens are using Open Government Data and the benefits they expect open data to bring to society. The full report was published on October 19, 2017. The study had some interesting findings, such as: nearly 8 in 10 respondents say open government data can improve the lives of citizens; 25 percent of people in Singapore are using open government data to create new businesses, which is comparable to South Korea and India at 30 percent; and Singaporeans generally trust their government to keep their data safe.

You may also like

Show Buttons
Hide Buttons