Why using ads and boosted content to promote health information is problematic

One of the easiest ways of studying messages and user behavior on the internet platforms is to leverage platform ads or “boosted content” that has been paid for. This is how the platforms monetize their user data because marketing professionals can use it to real-time optimize their marketing campaigns.

But how much can it help achieve a health authority’s goal to reach all and especially vulnerable communities with health information?

Health information equity cannot be achieved through boosted content

If we want to achieve true resilience of individuals and communities to misinformation, we need to ensure that they all receive credible and accurate health information in ways they can understand and use it.

• Boosted content is usually geared toward a specific audience that the marketer is trying to reach. Ads to promote campaigns inherently run for only certain segments of populations and not everyone.
• This can leave people and communities out of the digital efforts to promote acceptable, relevant and credible content dissemination online.
• While relying on ads to boost content may make sense for purposes of marketing, this cannot be the dominant strategy for health information dissemination in the digital spaces.

Moreover, analytics on boosted content is not available to those that are not account owners on the marketing dashboards of internet platforms.

• For example, Crowdtangle does not include information about what boosted content on a specific topic is served to a particular audience.
• This basically means that what little public data is available on what is being discussed and shared about a given topic is utterly missing information on what content has been boosted – and this content actually most likely has the farthest reach and engagement, and therefore can influence the dynamics of the larger conversation.
• Because of this, we cannot account for the displacement effect of boosted content because we do not even know the share of voice it has compared to non-boosted content.

All this said, imagine trying to understand discussions about healthy eating and how people talk about dieting – but we do not know how much content is being served that is bought and paid for by pharmaceutical companies, and for-profit companies selling diets and exercise plans, or even what boosted content is being served by influencers, who are paid by companies to market to a large audience.

If we are saying that we are trying to ascertain the drivers of conversations online, we by design are cut out from understanding how individual marketing accounts are boosting content through their campaigns and how this interacts with the conversations online.

Ad-driven research on platforms

Another issue related to ads on social media and internet platforms is that the platforms use their ad provisioning mechanisms to offer free ad credits to international institutions like the UN or other nonprofit, government or academic organizations (see example here and here).

These ad credits can be used by platform clients for common optimization of marketing campaigns that are sending messages out to people. For example, Facebook has a business service that helps their clients to promote their brand on the platform. By providing ad credits, it engages not-for-profit, health and government sectors in activities to promote their brands and advertise effectively on the platform.

When a marketer uses these ads, they pair them up with tools that basically allow them to at scale test what content performs better in what segments of users. After these tests, the best-performing are promoted further and this allows the marketer to do a better job pushing their messages to the target audiences. This can be used also in health to optimize delivery of messages on health topics to audiences online and encourage action can be tracked through click-through rates.

The difference between a company using social media ads for marketing and a public health authority is that the company wants to sell you a product which is why they want you to click and shop. Their ultimate metric is number of sales generated. For a health authority, it may be a click-through to a web page with more health information, or to book a vaccine appointment.

Therefore, a health authority’s ultimate goal with an ad credit buy beyond building awareness might be a larger variety of actions that a marketing approach to metrics just won’t cover. As far as I know, no one has been able to link social media ads to an increase in vaccine uptake by linking exposure to ads online to offline behaviors like getting vaccinated.

In addition, these ads can also be used to ask users to contribute to surveys or to understand people’s knowledge about a particular health topic. Facebook ran ad-driven surveys and other studies in collaboration with American Universities during COVID-19 pandemic through their Data for Good programme. The utility of such datasets is limited for anyone doing epidemiological analysis to understand specific geographic communities.

A nerd’s frustration with the usefulness of brand lift studies for the promotion of health behaviors

A lot of research has been conducted during COVID-19 trying to determine the link between information sharing, amplification and exposure online to how it influences offline behavior. The chain between exposure, feeling and acting is an area that behavioral scientists have studied for decades.

What’s novel about the information environment that we are living in now is that we have far more access to data on how a piece of misinformation is shared and amplified and who is exposed to it.

However, online platforms have monetised access to this type of information for marketing purposes to help sell products or sway audiences to a particular political perspective. Unfortunately, there isn’t always a separate pipeline for access to this type of data for public health research purposes (that i discussed in part 1 of this thought exercise).

Limited approaches such as brand lift studies have become more popular in the past few years, but generally speaking they have pretty modest influence on people’s perceptions and intent to perform a particular health behavior (see Athey S, Grabarz K, Luca M, Wernerfelt NC. The effectiveness of digital interventions on COVID-19 attitudes and beliefs. National Bureau of Economic Research; 2022 Jul 25.)

Online platforms have been providing a lot of ad credits to different health and social welfare-focused organizations specifically to provide free advertising or utilize brand lift studies for research.

Building on top of platforms designed for digital advertising and serving limited data on how users interact with these ads doesn’t incentivize the platforms to provide an alternative and a more robust data stream that can be useful for public health research and action, especially during emergencies.

Currently, we have researchers searching for ways to access the data, and platforms having to navigate the implementation of their own changing data privacy and content moderation policies. There clearly is a need for responsible processing, access and use of this data. This is an analogous need for access to personally identifiable information for the public health interest that I mentioned in part 1 article here.

I haven’t seen this mentioned in the discussion of platform regulation of data privacy and access legislation. If anyone reading this has more information or references to share on this, I would be interested to receive them.

Because of all these limitations of the online platform datasets, we’ve been emphasizing the need for the established evidence-generation practice in health – triangulation of data sources (of varying quality, limitations and coverage) for the purpose of understanding questions of interest for a health programme. Our manual on how to produce infodemic insights is one such tool.

Voluntary promotion of content during an emergency

During the pandemic, platforms voluntarily promoted content about COVID-19 resources and information from government websites. When a user was searching for or reading COVID-19 content on a platform, they would be offered links to sources of health information in their country.

This was a powerful way to channel users to health authority websites, but it’s unclear how well the health authority websites were segmented and answered specific questions and concerns that were circulating. For example, Google posted help pages for how to prepare government web sites to serve up-to-date COVID-19 content to the search engine. It would be useful to know how many government websites took the recommendations into consideration.

In addition, there is no easy way to go back and analyze what links were promoted, for how long, to whom, and with what impact. Additionally, the process by which the platforms determined what was considered authoritative would need to be transparent in the future. When such an arrangement is put in place, any weaknesses in the currency of the content, or serving it digitally in ways that is easily syndicated become more visible.

What are the standards to adhere to when governments, professional associations and other formal bodies set up digital content and websites so that their content is up-to-date, SEO optimized, syndicated and machine-readable when it is indexed in search engines or boosted? These should exist and public institutions could be benchmarked on the performance of their websites in the information environment.

PS: An epidemiologist’s approach to understanding information exposure and health behaviors

We have overwhelming evidence to know that exposure to health information does not easily translate into health behaviors. (see this article from Carrasco-Farre in Nature unpacking some nuances). So how can one effectively research that link?

Thinking like marketing professionals won’t get us closer to improving the routine datasets or methods to measure essential indicators needed to inform infodemic management. Thinking like epidemiologists will.

An example is a project in which the WHO infodemic management team has been working with the University of Sydney on information exposure measurement tools that focus on harmonized measures of linking information exposure and health outcomes at the individual level, analogous to other population-based epidemiological studies.

We need more thoughtful research linking our efforts to influence information exposure of people to individual outcomes and ultimately population effects.

I wrote this LinkedIn blog in early 2023, to point out some drawbacks on using communications campaigns on social media platofrms as the primary action in digital spaces to promote health information. Follow me on LinkedIn. if you’d like to read more of my commentaries.