Possible Topics

COVID AND MISINFORMATION

EXAMPLE SCENARIO: ASYMPTOMATIC AND SURFACE SPREAD

Among the things the general public and government officials now know is that asymptomatic individuals can transmit the virus, and that hard surfaces (among others) can hold the virus for later transmission. This misunderstanding of transmissibility arguably kept WHO from declaring a pandemic until March. It also led to public officials making blatantly inaccurate statements or engaging in hazardous conduct (e.g., drinking in crowded bars) into March. In practical terms, COVID-19 is widespread enough that every person should treat themselves, surfaces, and others as possibly infected.

Challenge:​ Create models to track, on a weekly basis, news articles or official pronouncements over the evolving understanding of disease transmission. Network nodes could include assurances and warnings about transmission.

Why this matters:​ Inaccurate understandings of COVID-19 transmissibility may have dramatically accelerated expansion of the pandemic, and remaining misunderstandings will undermine recovery.

Alternatively:

• What types of misinformation about COVID-19 are being spread on Twitter (or other social media platforms)?

• What are the general themes of the COVID-19 tweets by news media and government/international organizations?

• How are particular forms of COVID-19 information spread on Twitter related to the spread of the virus/number of deaths/government actions?

Possible qualitative datasets

Dataset3: English news articles

Dataset5: Automated Twitter accounts

Dataset6: News media and government/international organization tweets

Possible codes / frameworks

Coding scheme for media coverage and the role of the experts (Vasterman & Ruigrok, 2013):

Alarming: Negative descriptions of the virus; Negative consequences; Negative pragmatic markers (warnings); Negative perlocutionary force markers

Reassuring: Positive terms for the virus (mild virus, etc.); Putting risk in perspective of other risks; Positive consequences; Positive pragmatic markers; Positive perlocutionary force markers

COVID AND MEDIA REACTION

EXAMPLE SCENARIO: CONFIRMED CASES VERSUS ACTUAL SPREAD

The daily count of new confirmed cases is arguably a function of two major variables: infection spread and testing capacity. It significantly underestimates actual prevalence. Yet increases in the confirmed cases seem repeatedly taken as indications of spread alone. This is highly misleading, in that it treats confirmed cases as a proxy for spread, badly underestimating spread, and is an inherently incorrect value for the rate of spread.

Challenge:​ Create models to track co-occurrence of news articles relaying confirmed case counts with qualifying statements about actual count, about confirmed case counts as low estimate, clarity about actual spread as affected by testing capacity, direct statement or inferring that confirmed cases is the indicator of spread.

Why this matters:​ Public understanding of the nature of disease transmission rates is critical. When confirmed cases are low, public attention is slow. During the current period of early April, a daily increase in the US that is constantly fluctuating in the 25-35K range likely reflects maximum testing capacity rather than actual spread. There are arguably 1.5m or more cases in the US. Let’s say by April 30, the figure is 1.6 million, reflecting mitigation and recovery. Over a three week period, 600,000 confirmed cases are added. The public treats this as increasing spread at a doubling rate rather than a 7% increase. The confirmed case count drastically understates what the general public should be aware of.

Alternatively:

• How does the context of neews articles change over time?

• What sorts of COVID-19 fears are most common, and how do they change over time?

• How does the CDC information provided change over time as the virus progressed/spread? How is the public responding to CDC tweets? What about other national health agencies?

Possible qualitative datasets

Dataset3: English news articles

Dataset6: News media and government/international organization tweets

Possible codes / frameworks

Panic, testing, economic effects, social effects, recovery, death, medical updates, lies/government cover-ups, statistics

From Jung et al. 2012 to code news: Dominant news frame-new evidence; Attribution of responsibility; Uncertainty; Reassurance; Consequence; Action; Bare statistics (no frame)

From Lee & Basnyat, 2013 to code press releases and news: Basic information; Preventive information; Treatment information; Medical research; Social context; Economic context; Political context; Personal stories (experiences of patients/families); Other (open-ended)

REACTION TO COVID CONTAINMENT AND MITIGATION MEASURES

EXAMPLE SCENARIO 1: FACE MASKS

The WHO has repeatedly noted that there is no proven effect of the value of non-medical grade face masks. Yet some countries with notable success in mitigating viral spread practice face masking nearly ubiquitously (e.g., Korea, Japan, Hong Kong, Singapore), though there are exceptions to their success. Different countries are now adopting either strong recommendations or requirements for face masks. It is quite possible that with more accurate understanding of transmissibility and ubiquitous use of face masks in Iran, Europe, and US, the pandemic would be much less severe. This scenario does not seek to answer that question, but does seek to model policy recommendations and requirements concerning what is likely to become widespread practice.

Challenge:​ Create models to track evolution of discourse on face masks, with nodes representing issues such as concerns of availability for health workers, lack of gold-standard evidence of effectiveness for non-medical grade masks, WHO recommendations, individual country recommendations (e.g., CDC), types of face masks, and expected efficacy.

Why this matters:​ Public understanding of each individual’s agency and responsibility in both personal protection and protection of others will be a crucial part of the recovery process.

EXAMPLE SCENARIO 2: POLITICAL UNITY

A common expression in times of crisis is that crisis unifies a population. However, political leaders in different countries have taken very different approaches to achieving unity.

Challenge:​ Create models to track the evolution of the tone of discourse in different countries, with nodes representing both specific and general factors (policies on closings, ventilator availability/access, etc.) and something like the collaborative or cooperative valence.

Why this matters:​ Despite some successes in people working across partisan divides, leadership in many countries has simply failed to rise above acrimony when collaboration and joint problem-solving is necessary. Moreover, while a sense of collectivism has arisen in many locations, nativism, racism, and authoritarianism have increased in others.

Alternatively:

• What are people’s main concerns with the spread of the virus and how are these concerns related to statistics such as the spread of the virus, number of deaths, number of tests, and governmental actions?

• How does sentiment (concerning COVID-19) change over time and in relation to containment measures? What types of messages are connected to which type of sentiment? How do these differ regionally?

Possible qualitative datasets

Dataset1: English online discussions

Dataset2: English blog posts

Dataset4: COVID-19 tweets dataset

Possible codes / frameworks

Sentiment (pos/neg/neutral), tweet content (perceived threat serious or not serious, hygiene measures necessary or not necessary, adherence or non-adherence to containment measures), date (as discourse segmentation), location (as discourse segmentation or metadata), containment measures (as metadata)

From Chew & Eysenbach, 2010: Humour/sarcasm; Relief; Downplayed risk; Concern; Frustration; Misinformation; Question

From Signorini, Segre, & Polgreen, 2011: The volume of Hand-Hygiene- and Mask-Related tweets; Travel- and Consumption-Related tweets; Drug-Related tweets and Vaccination tweets over time