Bias and social justice in AI contact tracing apps developed during COVID-19 pandemia: a scoping review

During the COVID-19 pandemic, one of the most recurrent measures developed to control and mitigate the impact of the virus has been the development of digital contact tracing technology based on artificial intelligence (AI). Contract tracing apps were established as a new public-health intervention in many countries during 2020 along with other preventive measures to help slow down the chains of transmission. Through self-reported information of symptoms and geolocalization they were intended to find and advise possible close contacts. Through this scoping review our aim is to analyse what are the biases identified in these AI systems or algorithms in the scientific literature and their relation to Social Determinants of Health (SDOH). We carried out the search strategy in Pubmed, Medline, CINAHL, Scopus, Wiley Online Library, WOS, and Arxiv.org. We have identified biases related to: 1) inadequate data collection (underrepresented clusters, self-reported data, unrepresentative training data, false positives or asymptomatics), 2) technological biases (access to smartphones, lack of infrastructures, poor data connectivity, etc.), 3) SDOH (age, race and socioeconomic status information). Moreover, there are many concerns regarding possible violation of security, privacy and data availability of COVID-19 patients. Even if great efforts have been done to design and implement digital contact tracing AI during the outbreak of COVID-19, there are no systematic studies on either the effectiveness of their use as a public-health measure or their fairness regarding possible biases in their design and application that can lead to harmful effects on the most structural disadvantaged groups’ health.