For the third year Edelman has released their health special edition of their famed Trust report. In this edition they tackle trust in third parties, healthcare companies, the media, and AI. 

As always, the data collection is impressive. The researchers surveyed 15,000 respondents over 16 countries.

Some things never change and in this case only get stronger. The family doctor is still the most trusted advisor for health decisions followed by pharmacists and friends and family.

This one makes the cut because the US dropped two rankings ... from just barely trusted to just barely distrusted here. The question was "I trust healthcare companies to do what is right." The industry might have some work to do here.

Media outlets have been dropping in trust for years as the populace gets more divided and outlets are perceived to have biases. Note, however, that Edelman is now showing the change number from 2019, a much longer time than above (which was one year).

That's not good. Technology, when used correctly, absolutely has the ability to improve healthcare. Perhaps that is the exact issue though, patients don't trust the people building the technology (speculation, not in the report -Ed.).

In any event, people claim they don't want chatbots for medical communication. With the usefulness of chatbots before AI it's not surprising. Let's see if AI can change their minds in next year's report.

These reports are always a big hit in the industry and this one looks to be just as successful. It has 85 pages of data and charts and even the appendix material is interesting to folks who want to use ChatGPT to find some hidden gems in the numbers.

50% of pharma & biotech companies restrict the use of ChatGPT by employees because of data exposure risks. That number bumps up to 65% when limited to the top 20 large pharma companies.

ZoomRx surveyed 211 industry employees who were aware of AI (which was likely all of them anyway) and across all job types.

Endpoints seems to have the scoop on this story, we couldn't find the study anywhere public.

Google has released a prompt-writing guide for Gemini, it's AI tool integrated into it's search, docs, email, and other products. This guide is a great introduction to prompt writing for new users and a good reminder for experienced ones.

The recommended prompt length is 21 words, give or take, which will seem long for new folks and incredibly short for many experienced users. Google breaks this down into four segments, common in prompt guides:

The document has extensive examples for six business users:

A good side effect of the examples is that they highlight what AI is not good at. They show use cases of helping organize ideas while showing the thinking that goes into the prompts. 

Finally, the authors end with some style advice for good prompts (these are all quoted from the document):

For that last one folks at Klick have also advised to simply tell the tool to "try harder." That works a surprising number of times. 

The IAB, with its partner PwC, have released the US 2023 Internet Advertising Revenue Report full year results. The trend is up with search still firmly in first place.

The IAB doesn't break social out very often because it seems to include it across different media. We say "seems" because it doesn't address this measurement directly so we're left to guess.

Growth by percentages slowed into 2023 by IAB segmentation.

However, growth was up when looking just at social media.

the 2021 growth mirrors search as these two channels were the easiest to adapt to the changing environment due to COVID-19 lockdowns and restrictions.

Finally, let's look at digital compared to other channels. To say digital dominates would be an understatement.

Grab a copy to see the full IAB analysis, just don't expect any detail about social.

Two groups at Stanford have collaborated on an AI opus that summarizes much of the industry at this point in time. Because we're talking about AI this document will go out of date quickly, but for now it is the most comprehensive overview we've found. At 502 pages we won't even try to summarize it, but we will highlight three things pharma marketers may care about. 

Two groups at the University of Toronto are releasing a new report in May about global opinions about AI. This report seems to have a scoop on that data. One of the charts shows how often people familiar with ChatGPT use the tool. All we can say is that usage is high. No country is over 50% for "rarely."

India, with its robust information technology industry, is highest in terms of usage frequency. However, keep in mind this is data for ChatGPT users and doesn't account for non-users. The full report is due in May.

This evaluation of medical knowledge has been used since 2019 and at that time models were around 40% accurate. Now with the latest GPT-4 test, that LLM has achieved 90.2% accuracy.

Comparing a few different models against four different tests we see it's still a tight race in medical knowledge accuracy. GPT-4 Medprompt from Microsoft gets a slight edge over the competition across the board.

An application that enhances pandemic prediction, offering pharmaceutical marketers working on vaccines a tool to better forecast and manage public health campaigns.

For the other 499 pages in the report, grab a copy today, especially since it might be out of date in a month (exaggerated, but directional, so I'll allow it -Ed.).

In its "State of Design & Make" report, Autodesk surveyed 5,399 professionals across 17 industries. 

When asked if they thought AI was a threat to their industry and company the respondents were evenly split.

9% of the D&M group was in Life sciences manufacturing which is still 161 respondents. Also, 42% of the M&A group was advertising respondents which makes that group 663 people. Since marketing in life science is our focus we combined a couple of charts to show how many in each group are using AI for different purposes.

The actual question was what are the use-cases for AI in your company so these answers include both what's currently happening and what the respondents can see in the near future.

Autodesk is one of the original 3D development environments and was founded in 1982. This company has deep roots in many industries that use 3D modelling and this report is worth a look for those wanting to understand the zeitgeist of the workforce.

Generative AI models aren't as persuasive as a human argument... yet. Anthropic wanted to see how well its latest LLMs could craft persuasive arguments. It turns out they are getting better over time and approaching human-like abilities. 

The study had participants rate their agreement with a claim, then were presented with an argument to get them to agree with the claim, then they re-rated their agreement. The humans used for comparison were not professional persuaders but they were competent communicators. 

The statements and argument lists can be accessed or downloaded as part of the full dataset from Huggingface. We took a quick look at it with ChatGPT (sorry, Anthropic) and it seems like 12% of all the arguments touched on health topics.

Perhaps not surprisingly the people with opinions in the middle of the range were more likely to adjust their opinions than those at the extremes.

With very few negative numbers in the changed ratings we can see that the AI models have the potential to change public perception. For marketing this is essentially benign but we do want to continue to monitor for other, less up-front, uses of the technology.

The number of online reviews of healthcare providers has grown from 643K to 1.6M annually. The overall ratings haven't changed much, but they are getting more positive which Reputation attributes to providers asking for reviews more often.

Here, we see that the response rate has increased which Reputation attributes to providers engaging with their patients.

The firm highlights the importance of these ratings from a previous study conducted in 2023:

LinkedIn has released a new "premium" company page format that features AI-assisted post writing assistance. Of course, you could just use ChatGPT for that but it's possible that the social platform will have enough data on the brand that it will produce better results.

Some other features that come with the $100 / month subscription are:

Moves a CTA from the drop down menu to the company info bar.

Incorporates a 40-pixel image and 80 characters of text. Less than an X-post, make it count.

This one could be interesting. Members who engage with the page content will be auto-invited to follow the company. Hopefully this doesn't train LinkedIn users to not engage as much.

A gold IN logo on the Page header. Because, of course. 

A children's cancer non-profit was facing a challenge, a steep 42% decline in donations in 2022. To combat this decline the team created the League Of Undercover Donors (LOUD).

The campaign consisted of 300 heavy, metal coins with a QR code on the back. Scanning the code took the user to the donation site and a reversal in messaging for a charity:

This gamified campaign had impressive results:

Last week CDER published information on its CDER Center for Clinical Trial Innovation (C3TI). Which really does sound like a Star Wars droid name (stay focused -Ed.).

In the program description, a couple of links in, are three demonstration projects that showcase the types of trial innovations the agency is looking for.

Anyone designing their next clinical trials may want to pay attention to this.

Data blocking is a longstanding issue in the US health system but is being reduced year by year. So, when we saw this story we were initially skeptical of Epic's decision to turn off data to another company, Particle.

However, on the surface the reasoning seems valid:

A very long but detailed and objective review of the issue is available from the Health API Guy. Anyone interested in HIPAA and various data blocking initiatives designed to force the large players like Epic to make their data available will be interested in this saga.