[Paper Review] COVID-19 and the Social Distancing Paradox: dangers and solutions

Massimo Marchiori|arXiv (Cornell University)|May 26, 2020

COVID-19 epidemiological studies19 references49 citations

TL;DR

The paper presents the world-first sensor-based dataset on social distancing during COVID-19 in Italy, showing a paradoxical clustering at short distances when unmasked and a distance boost when masks/goggles are used, with weak policy impact.

ABSTRACT

Background: Without proven effect treatments and vaccines, Social Distancing is the key protection factor against COVID-19. Social distancing alone should have been enough to protect again the virus, yet things have gone very differently, with a big mismatch between theory and practice. What are the reasons? A big problem is that there is no actual social distancing data, and the corresponding people behavior in a pandemic is unknown. We collect the world-first dataset on social distancing during the COVID-19 outbreak, so to see for the first time how people really implement social distancing, identify dangers of the current situation, and find solutions against this and future pandemics. Methods: Using a sensor-based social distancing belt we collected social distance data from people in Italy for over two months during the most critical COVID-19 outbreak. Additionally, we investigated if and how wearing various Personal Protection Equipment, like masks, influences social distancing. Results: Without masks, people adopt a counter-intuitively dangerous strategy, a paradox that could explain the relative lack of effectiveness of social distancing. Using masks radically changes the situation, breaking the paradoxical behavior and leading to a safe social distance behavior. In shortage of masks, DIY (Do It Yourself) masks can also be used: even without filtering protection, they provide social distancing protection. Goggles should be recommended for general use, as they give an extra powerful safety boost. Generic Public Health policies and media campaigns do not work well on social distancing: explicit focus on the behavioral problems of necessary mobility are needed.

Motivation & Objective

• Identify why social distancing under pandemic conditions diverges from theoretical expectations.
• Create and deploy a sensor-based dataset to capture real-world social distancing behavior.
• Analyze how PPE (masks, DIY masks, goggles) affects distancing behavior.
• Assess the impact of public health campaigns and regulations on actual social distancing behavior.
• Propose actionable strategies to improve safety in future pandemics based on behavioral insights.

Proposed method

• Developed a hidden sensor belt with ultrasonic distance sensors to measure real-time social distances.
• Collected data in Venice metropolitan area over 2+ months (Feb 24–Apr 29, 2020) across sidewalks of widths 163 cm, 175 cm, and 222 cm.
• Evaluated five scenarios: unmasked, masked, DIY-masked, goggles+masked, goggles+DIY-masked.
• Programmed the microcontroller for temporal reasoning to handle cases like quick passes and to avoid false readings.
• Compared behavior across national and regional policy changes (N1–N3, M, R1–R3) and analyzed temporal evolution.
• Presented distributional analyses of social distances and how PPE alters the distribution shape.

Experimental results

Research questions

• RQ1 What is the actual distribution of social distances in pandemic walking scenarios without masks?
• RQ2 How does wearing masks, DIY masks, or goggles alter social distancing behavior?
• RQ3 Do public health mandates and media campaigns measurably shift real-world distancing?
• RQ4 Does PPE trigger a shift from paradoxical to safer distancing, and by how much?
• RQ5 Are there consistent effects across sidewalk widths (163 cm, 175 cm, 222 cm) and over time?

Key findings

• Without masks, the average social distance is 29.4 cm, with a paradoxical distribution skewed toward the other person.
• Wearing a mask increases the average distance to 58.42 cm and can extend the distribution beyond the sidewalk’s max width (e.g., 119.5 cm).
• DIY masks raise the average distance to 69.02 cm and extend the distribution up to over 150 cm.
• Goggles with masks raise the average distance to 79.79 cm, with further boosts when combined with DIY masks (e.g., 92.39 cm at 163 cm sidewalks).
• Goggles provide an additional safety boost beyond masks alone, consistently across sidewalk widths.
• National/regional policies and media campaigns showed no statistically meaningful change in social distancing distributions (p-values > 0.59) across cases.