Getting ahead of the curve in Singapore with mobile phone metadata
Written with Zhihao (Z) Lin
It is now the second month of our Circuit Breaker and the costs of such a lockdown have become increasingly apparent. We should consider all options to minimize this expensive disruption and do our utmost to prevent resurgences.
The combination of personal costs (lost jobs and income, emotional toil) and the widespread economic impact (entire industries disrupted, cascading delays due to work stoppage or reduced productivity) makes it impossible to fully quantify the costs of each day of our lockdown, but there have been some attempts by economists.
One estimate puts the net loss for the US economy due to the shutdown of “non-essential” businesses at $28.2 billion per day, and “… estimated to total almost $15,000 per household per quarter….” It goes on to say “this is why better techniques for fighting the war are incredibly valuable”. We are not aware of any similar studies for Singapore but would imagine that the impact would be of a similar magnitude.
There is much to be done to manage the pandemic and ensure economic recovery. One technological approach that has been gaining popularity is the use of Bluetooth-based contact tracing apps. The biggest stumbling block to the app approach (Bluetooth or otherwise) is adoption, a challenge also mentioned in the TraceTogether whitepaper.
In Singapore, the TraceTogether app has been downloaded by only about 20% of the population, falling short of the 60% needed, according to Oxford researchers. We also seem to be falling back to more manual techniques such as collecting NRIC numbers when entering crowded markets. Beyond the exhortation by PM Lee for everyone to install the app, there are suggestions to make TraceTogether mandatory or highly incentivise its use, but if it is mandatory and it is so crucial to have more comprehensive coverage, then why limit ourselves to app-based systems?
Using mobile phone metadata
We propose that mobile phone location metadata be considered, in conjunction with other solutions, to help us move more quickly out of the Circuit Breaker and keep us safe. Early in the crisis, there was talk about using mobile phone metadata to help understand the movement of people and facilitate contact tracing. Success stories come from the likes of Taiwan, South Korea, and Israel.
Mobile phone metadata is already being collected and used by the telcos for operational purposes such as improving network coverage and fixing dropped calls. It works 24x7 with no installation, as long as the phone is turned on and connected to the network. It works on iPhones, and Android phones such as those from Huawei, Samsung, and Xiaomi, and will work even on older feature phones, which may be used by our foreign workers, the poor, and the older generation, none of whom should be excluded from our COVID-19 response. It will also work with visitors from other countries who choose to roam with their existing SIM cards rather than buy a local prepaid one.
Why the loss of interest in mobile phone metadata? The biggest question has been privacy, but we believe the privacy concerns of mobile phone metadata can be addressed, and it will be more acceptable and useful than other alternatives such as making app usage mandatory. Privacy concerns are also not specific to using mobile phone metadata, but applicable to all contact tracing approaches.
Privacy protections
There are a few ways of tackling the privacy issue:
- Choice: Ability for people to opt-out.
- Transparency: Clear visibility for individuals to see their data collected and who will have access to it.
- Technological barriers: Encryption and data architecture to limit identifying individuals (by phone number or NRIC) unless necessary.
- Regulations: Laws to prevent the use of the data by anyone except the MOH, and even then only to address the current COVID-19 pandemic.
Having a choice is critical to our freedom, but making things opt-out rather than opt-in greatly helps with adoption and there are precedents in Singapore for critical needs such as organ donation. Those who opt-out may have to resort to inconvenient options such as presenting their NRIC to enter shopping malls.
Consent has to be well-informed to be valid. In this case, informed consent means knowing what data is being collected and the “chain of custody” of which intermediaries will handle the data and who is finally using it. TraceTogether does a good job here, including open-sourcing their code.
The telcos already collect the mobile phone metadata, so it is only necessary to limit how much data is shared between the telcos and with the government. There are various alternatives involving some clever encryption and architecture to minimise the exposure of the individual’s locations until it is necessary to address the spread of infections.
Australia and Israel are enacting laws to limit the use of data from contact tracing apps to health authorities. A similar law in Singapore could be applied to app data and other data sources and is a clear covenant between the government and citizens about how the data will be used and for how long the data will be stored: a “sunset date” of no more than 14 days which is more than enough for COVID-19 contact tracing. It is a stronger commitment than just an end-user license agreement (EULA) from an app.
The other concern about using mobile phone metadata is its precision. It is less precise than Bluetooth- or GPS-based solutions but even such solutions need a “human-in-the-loop” (manual reviews) to minimize false positives. The best approach would be to bring all solutions together and leverage the precision of Bluetooth contact tracing apps with the ubiquity of mobile phone metadata, plus good human commonsense. We had success with “live” trials elsewhere using only mobile phone metadata through the combination of machine learning techniques and “human-in-the-loop”.
There are other ways mobile phone metadata can be used as part of the pandemic response. Taiwan uses it to help enforce home-based quarantines. The data could be used to identify at-risk populations before a serious outbreak occurs, a scenario that Singapore is now all too familiar with.
Mobile phone metadata solves the key challenge facing contact tracing mobile apps today — adoption. We believe that privacy concerns can be mitigated by providing choices, implementing technological barriers, passing strict laws, and by providing transparency above all. Ultimately, every country will have to balance between the adoption and utility of its contact tracing solution versus the privacy concerns of its citizens. As South Korea “…had laws revised to prioritize social security over individual privacy at times of infectious disease crises”, East Asian societies have leaned towards the good of society over individuals and this has seen broad support from their populace.
Who are we?
We are a team of Singaporean data scientists and experts in mobility analytics but we are not epidemiologists or public health experts. Together, we have decades of experience working with mobile phone and Wi-Fi location metadata, and with the telcos to provide analysis on how people move in cities. Our opinions here are through the data analytics lens.
