AI in Times of Pandemic: Between Hope and Concerns

AI has played a substantial role in assisting public health and the medical community to curb the current pandemic. However, are AI contact tracing apps really useful? Will they harm individual privacy and liberty?

By Hubert Laferrière

Over the past weeks, media have been reporting on the development of digital applications that would make it possible to monitor Coronavirus transmission chains by collecting travel and interaction data from individuals using AI technology. These apps, known as peer-to-peer contact tracing apps, are fuelling discussions on the issue of individual privacy and liberty. Attitudes of circumspection and mistrust are displayed regarding the potential effects of these apps when used in the field of public health. Some are raising concerns about potential harmful side effects of this technology on the democratic institutions and culture, including freedom of choice and movement, some are qualifying the apps as “invasive surveillance measures”. [1]

AI is Stepping up against Covid-19

There is however no doubt that AI is and will play a substantial role in assisting the medical community to curb the current pandemic and find a cure. For example, an AI technology has identified preliminary signs of the pandemic. On December 30, 2019, BlueDot, a Canadian start-up firm, using machine learning (ML) to monitor outbreaks of infectious diseases around the world, alerted various governments, hospitals, and businesses to an unusual bump in pneumonia cases in Wuhan, China. Nine days later, the World Health Organization officially flagged what we have all come to know as COVID-19. [2]

Last year, the U.S. National Institutes of Health (NIH) confirmed that “Machine Intelligence (MI) is rapidly becoming an important approach across biomedical discovery, clinical research, medical diagnostics/devices, and precision medicine.” [3] Experts are now using MI to study the SARS-COV-2, test potential treatments, diagnose individuals, analyze the public health impacts, detect early signs of virus explosion, monitor public safety, and more.

Last February, scientists at the Massachusetts Institute of Technology and Harvard used a deep learning neural network to identify a molecular compound different from most antibiotics, that was discovered from the Drug Repurposing Hub and demonstrated it’s effectiveness against a bacterium that no existing drug can eliminate. The discovery has implications for fighting the SARS-CoV-2: finding a novel use for an existing compound, and using neural nets in place of familiar chemical definitions.[4] 

Scientists from Carnegie Mellon considered an early version of COVID Voice Detector, an app that would analyze a patient’s voice to detect an infection. There is an “AI powered stethoscope” (Eko) helping doctors to directly treat patients with “wireless auscultation” of the heart and/or lungs. This is helpful when practitioners wear significant amounts of protective gear. Medicago, a Québec city-based private firm, is already working on a vaccine using ML algorithms to measure 10,000 to 15,000 molecules. In a nutshell, AI initiatives and activities are buzzing in life sciences and technology areas.

AI and COVID -19 Contact Tracing 

The use of contact tracing applications aims at supporting the public health authorities by accelerating their efforts to target the transmission chain of the Coronavirus and individuals by assisting them to better manage social distancing, particularly useful with the lift of lockdown measures. 

MIT Technological Review has documented 25 individual, significant automated contact tracing efforts globally (as of May 7).[5] Some countries, mostly using proposed solutions developed by Google and Apple, have indeed deployed the applications (France, United Kingdom, Germany, China, Taiwan, Singapore, Australia, New-Zealand, Switzerland, Israel, etc.). Interestingly, these apps are essentially government mandated, whether their use is imposed or voluntary.

Tracing applications seem to trigger a high level of concerns regarding privacy and personal information protection including the management of data. Most of the countries have stressed the importance of maximizing measures to protect personal information. However, people are concerned with the government tracking them as well. In France, the Parliament approved the tracking app which was launched on June 2. The responsible minister indicated that StopCovid is a tool at the service of citizens which does not represent a threat to their freedoms, emphasizing its “voluntary use”, “pseudonymization”, “temporary” duration and “transparency”.[6] 

Here in Canada, researchers at the Montreal Institute for Learning Algorithms (MILA) are designing a smartphone app for contact tracing, COVI is a peer-to-peer AI-based tracing app. “This could be an important ingredient in the fight,” said Joshua Bengio, Scientific Director at MILA, in an interview in Science Business.[7] But he recognized the use of the COVI to collect information may put in contradiction intersecting people’s trajectories and privacy. “The government and the public would not accept an app that doesn’t match our cultural expectations. I don’t think the government will force people to use something like this, so it has to be something people accept”.

At the provincial and federal levels, discussions are underway. The government of Alberta encourages its citizens to use the AB TraceTogether application, first used by the Singapore authorities. Some provinces may follow Alberta; the Québec government seems not ready to deploy such an application. Prime Minister Justin Trudeau says the federal government will soon be recommending to Canadians to download a COVID tracing app. [8]

Are AI Contact Tracing Apps Really Useful?

Singapore’s experience suggests the app is not a panacea and must first serve as a support for professionals assigned to the crucial contagion tracing task. A developer of the TraceTogether application has warned health authorities against any form of “technological triumphalism”.[9]   

In Iceland, where the app penetration rate in the population is among the highest in the world, a senior figure in the country’s COVID-19 response indicated the real impact of the app has been small, when compared to manual tracing techniques like phone calls. In this context, the technology appears to be close to useless and not the game changer everyone expected. [10] 

Chief public health officer Dr. Theresa Tam expressed concerns about the mistakes that could be made by contact tracing apps if not developed properly, particularly the chance of false positives, thus alarming unnecessarily a whole bunch of people.[11]  

Contact tracing apps will not be deployed in Belgium, where authorities prefer traditional methods derived from well-established public health practices. In the U.S., mandatory application-based contact tracing appears not to be on the table.

“Une fausse bonne idée”? [12]

Jason Millar, Canada Research Chair in the Ethical Engineering of Robotics and AI at the University of Ottawa, asserted in Policy Options [13] that a well-intentioned but unproven application might not produce the desired outcomes; it could make things worse. In addition to significant privacy concerns, Millar’s key concern is that contact tracing apps can reinforce existing social biases, thus stigmatizing locations and communities even if anonymizing measures are stronger. The apps still provide users with information about high-risk locations and therefore stigmatization may not be stopped. In South Korea, some members of the LGBT community provided false information for fear of being personally identified with traditional tracing practices. 

In commenting on the Alberta TraceTogether app deployment for Radio-Canada, Julie Paquette, professor of ethics at Saint-Paul University, has recognized that the Alberta version app is less intrusive than the Singapore original, but risks about privacy still exist. Will the data be removed from the application and erased since they were collected for a specific purpose? In addition to privacy, her underlying question appears crucial: What is the likelihood that the app will not be used for other purposes?[14] Paquette’s main concern is about the potential ongoing use of the apps outside the emergency context, triggering a shift in purpose in the intent: Will these exceptional measures shift to a new surveillance system and to a more general acceptance of this type of surveillance? Generalizing the use of tracing apps could lead to a fall sense of security and open the door to the development of a surveillance culture, as surveillance could become the norm and an integral component of social relations. 

In reaction to MILA’s COVI app, Jocelyn Maclure, President of the Québec Commission on ethics in science and technology, expressed his concerns about technological solutions applied to complex and persistent social problems, an approach he labelled as “techno-solutionism”. At the same time, he outlined that critics of the solutions tend to focus exclusively on the risks associated with the use of technology, setting aside some benefits or improvements generated by the solutions. He recognized that more sophisticated technological tools could be beneficial for the public health authorities but he called for robust measures to mitigate the ethical risks inherent in the deployment of such technologies: if it is up to experts in epidemiology, infectious diseases, and public health to define the needs in this area, it is up to independent ethics experts to address the ethical risks.[15] 

In fact, we understand little about how the tracing apps could affect society and human attitudes. While there is no doubt that AI contributes and will contribute to the efforts to tackle current issues generated by the pandemic, one can hope that some boundaries will be drawn during and after this pandemic nightmare.

  1. Fox Cahn, A., Domino, A., (2020-04-06), Tracking Everyone’s Whereabouts Won’t Stop COVID-19, Fast Company.
  2. Heaven, W.D., (2020-03-12), AI could help with the next pandemic—but not with this one, MIT Technology Review.
  3. Cutillo, C.M., Sharma, K.R., et alii, (2020-03-26), Machine intelligence in healthcare—perspectives on trustworthiness, explainability, usability, and transparency, Nature Research Journal- NPJ Digital Medecine
  4. Stokes, J.M., Yand. K., et alii, (2020 -02-20), A Deep Learning Approach to Antibiotic Discovery, Cell, Volume 180, Issue 4
  5. O’Neill, P.H., Ryan-Mosley, T., Johnson, B. (2020-05-07), A flood of coronavirus apps are tracking us. Now it’s time to keep track of them, MIT Technology Review.
  6. Laurent, Corinne (2020-05-28), Traçage numérique : l’application StopCovid validée par le Parlement, La Croix
  7. Kelly. É. (2020-04-02). Science in overdrive: Researchers are inspired and exhausted by scale of COVID-19 challenge, Science Business.
  8. Aiello, Rachel,(2020-05-22), PM Trudeau: Feds will soon ‘strongly’ recommend contact tracing app, CTV News. 
  9. Péloquin, T., Thibodeau, M., (2020-04-20), Géolocaliser la pandémie, une fausse bonne idée, La Presse. 
  10. Johnson, B., (2020-05-11), Nearly 40% of Icelanders are using a covid app—and it hasn’t helped much MIT Technology Review. 
  11. Boynton, John, (2020-05-03), Consent for coronavirus tracing apps must be ‘meaningful’, Canada’s privacy watchdog says, Global News. 
  12. Title borrowed from Péloquin, T., Thibodeau, M., op.cit. 
  13. Millar, J. (2020-04-15), Five ways a COVID-10 contact-tracing app could make things worse, Policy Options
  14. Seker, F., (2020-05-09), L’application AB TraceTogether téléchargée 140 000 fois depuis son lancement, Radio-Canada. 
  15. Maclure, Jocelyn, (2020-05-19), Entre le technosolutionnisme et le catastrophisme, La Presse

About The Author 

Hubert Laferrière 

Hubert was the Director of the Advanced Analytics Solution Centre (A2SC) at Immigration, Refugees and Citizenship Canada. He had established the A2SC for the Department of IRCC and led a major transformative project where advanced analytics and machine learning were used to augment and automate decision-making for key business processes.