1. General

Category

Category

Health

2. Project Details

Company or Institution

IV.AI in Partnership with 19 to Zero

Project

19 to Zero – COVID-19 Vaccine Hesitancy AI

General description of the AI solution

19 to Zero partnered with IV.AI to understand the noisy, unstructured conversations around the vaccine to reduce confusion, encourage safe behaviours, address misinformation, and empower health professionals to take action from a place of knowledge both online and in the real world based on all the conversations happening across the public internet and in the news. The AI output helped the larger partner network understand the core reasons people were hesitant about the vaccine. The layering of AI technologies enabled the team to understand hesitancy based on the nuances of each narrative while taking into account the emotion people were feeling about the complicated topic. As an independent nonprofit organization, 19 to Zero works closely with a broad, multidisciplinary group of partners. These organizations and individuals come from academia, healthcare, the public sector, private companies, and more to provide expertise across the full range of 19 to Zero’s activities. From polling and data analysis to healthcare worker outreach to mass communications and everything in between, this diverse coalition drives the work of 19 to Zero. The AI technology driving 19 to Zero takes a multimodal approach to understanding, engaging with, and shifting public perception towards COVID-19. Our team at IV.AI partnered with 19 to Zero to monitor and distil the conversations around the vaccine to encourage safe behaviours and vaccinations, address misinformation, and reducing the bias associated with public perception of COVID-19.

Website

https://www.19tozero.ca/who-we-are

https://iv.ai/

Organisation

IV.AI, 19 to Zero

3. Aspects

Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.

Our solution is live. We began working on the models when the first Pfizer vaccine was announced in November 2020. Since then, we’ve identified and are tracking between 1 – 2 million unique social graphs p/wk and monitor how conversations resonate to help inform health care workers engaging with various communities.

Our AI utilizes Deep Learning and Unsupervised NLP algorithms to understand the conversations happening on the internet more generally and isolate the ones in relation to vaccine hesitancy and how they resonate over time and across various demographics of the general population. We then look at the clusters that are generated that relate to the vaccine and split conversations based on pro-vaccine and vaccine hesitancy ideologies. The algorithms and software enable us to look at the structure of the conversations  to reduce human bias so we can better understand the intentsand nuances of people sharing on social media at scale.

We also track conversations based on sentiment to understand how various unique emotions such as fear and joy change in relation to each topic alongside more standard sentiment including positive, negative, and neutral sentiment changes over time. Tracking emotion responses alongside specific narrative understanding helps us measure how people are feeling about the different vaccine narratives to track how some concepts nudge emotional responses in different ways across different groups. Tracking emotion responses alongside specific narrative understanding helps us measure how people are feeling about the different public vaccine narratives alongside media, official, and government actions to track how some concepts nudge emotional responses in different ways across different groups and keep all those engaged in combating the virus informed with the most relevant and digestible insights.

Finally, we use community detection algorithms to identify natural social graphs to understand how different groups align around all the publicly available vaccine concepts.
The insights are shared with the public via social media posts from the Alberta Healthcare System, the CBC and other medical associations. The outputs are also being used to help train health care workers in Canada. Teams who are attempting to manage mis/dis information in order to save lives are relying on the outputs of our models so we also share results with the larger medical community and health care organizations. We very clearly track percentages of change with every report and use visualizations that help communicate the size and importance of the various conversations. This system creates a feedback loop where we learn new behaviours based on the change we enable.

Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.

COVID-19 pandemic has impacted virtually every human on the planet, posing significant risk to positive outcomes for many aspects of SDG 3 and 11 to be achieved. Our solution provides unique support to decision makers and healthcare workers in their attempts to influence a higher proportion of the population to get vaccinated. In achieving this outcome we are seeing an increase in the uptake of vaccination which has direct and in-direct impacts on both the health and well-being of citizens at an individual level but also a flow-on effect which enables the existence of safer and more sustainable cities.
The project has initially been aimed at the Canadian population, and we have been able to make a direct impact with the insights derived from the solution across the healthcare sector in this country. The main evidence of this being in our ability to measure millions of conversations happening on social media between Canadian citizens that show a significant shift in perception of, and trust in, the vaccine when compared to conversations happening prior to our solution being implemented.
We have the ability to roll out the solution across all UNESCO member states to understand and develop insights from any other social groups, that can then be used to nudge citizens toward positive behaviours en masse, in more than 167 languages.
However, the true potential of this solution goes beyond its current application. The AI developed can be directly applied to other use cases like climate change hesitancy, for one example, to make massive impact on many other global challenges that require behavioural change of large numbers of our population at the core of their solution. In extending the application of the AI, we build into the digital ecosystem for the planet – and the SDGs – a powerful tool for change.

Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.

We have been able to identify 2 million unique social conversations weekly to help us follow how groups are influenced and measure the pro-vaccine or vaccine hesitant positions. This data has helped the healthcare system in Canada communicate with its citizens to address their concerns about getting vaccinated based on related concepts we know that they care about and attempt to nudge their behaviour toward making a decision to become vaccinated whilst also helping the healthcare system manage dis/mis/mal information Canadian citizens are being exposed to . The social graph approach to monitoring conversation also allows us to identify underserved communities and share insights on how they are being influenced by different ideologies.

The solution enables a rich network effect by engaging national/international healthcare systems, universities, private sector organisations, minority groups and not-for-profits as well as community/faith based organisations.

Given the urgent response required, our solution received immediate uptake from the ecosystem mentioned above. In the broadest sense there have been 100s of millions of users engaged in our solution. This number takes into account the network of organisations and individuals involved in the development, execution and amplification of the solution as well as every citizen with a social media account who has engaged in a public conversation on the topic of covid-19. Given the international scope and social media data sources, adhering to and respecting industry standards and regulation such as GDPR has been vital to the success of the project. The project is built to scale from language to language to make the insights available to countries across the globe thus ensuring we maintain a collaborative approach.

Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.

Considering that the implications of the AI’s outputs can be so impactful, we have organized a large global working group to engage with the training and testing data of the AI models to help ensure we don’t overfit or misrepresent the concepts the AI is exposing. We also use an unsupervised approach to AI so that the models can find and expose natural occurring features that are happening in the raw data to ensure we consider human bias and expose the truth of the human activity that is considering the vaccine and influencing others who are following the conversation. As mentioned previously, we are also adhering to all regulation and best practise around the ethical use of data.

As mentioned previously, we are also adhering to all regulation and best practise around the ethical use of data.