AI: Good answers or bias confirmation?

So writers at Science.org, part of the American Association for the Advancement of Science, have found AI (at least the generalized large language models like ChatGPT) is not very good at summarizing scientific papers. While it was ok at summarizing scientific findings and commentary in layman’s language, it tended to sacrifice accuracy for simplicity and insert hyperbole (“groundbreaking” anyone?) where it wasn’t appropriate.

Seems it needed a lot of fact-checking as well as it couldn’t do things like differentiate between correlation and causation, or cite more than one result in multi-faceted studies. This led the folks at the Science Press Package team (SciPak) to conclude that ChatGPT didn’t meet their standards.

Now, the SciPak people were feeding ChatGPT specific papers and asking them to summarize to see if AI could do some of the heavy lifting real humans do when they report on science for the world at large. They had the ability to examine the papers themselves and see where ChatGPT was wandering off the path and re-prompt to get it back on.

But what are regular, intelligent, non-science folks who read about some study and decide to ask ChatGPT (or one of the other AI’s like Gemini, the AI behind its AI-mode search results) getting as answers? Curious, I did a (highly unscientific) test with the recent hot topic—Tylenol use during pregnancy—to find out.

Open to Interpretation

First, I asked the AI what the conclusions in the actual paper cited in the White House announcement (here on PubMed). Then I tried a general natural language query “Can Tylenol use during pregnancy cause ADHD or autism?” without referencing the study.

The results were interesting.

When given the paper, both ChatGPT and Gemini responded with a fairly accurate summary along the lines of “Maybe. There’s a correlation, but no definitive proof of causation. More study is needed.”

When fed the general query, the answer came back with “a causal link has not been established, but some studies have shown an association though these have not been proven…”

So the first answer is pretty accurate. “We’re not sure, but we’re looking into it.” The second sounds a lot more like “there’s no proof.”

Even more interesting is that the general query then returned all kinds of search results which were much more definitive in their “yes it does”, “no it doesn’t” answers, leaving it to the reader to make up their own mind as to what was correct.

Also, while the AIs cited high authority websites like Harvard, the FDA and Johns Hopkins, it also cited Reddit, YouTube, and Wikipedia along with them. (Note: I’m not a doctor, but I still recommend that you don’t get your medical advice from Reddit and YouTube.)

The markedly different answers led me to conclude that AI is pretty good at distilling information from a specific source, but left to do its own research it’s going to give people wildly different, probably inaccurate, answers. And, knowing people, they’re likely to pick answers that match their preconceived notions rather than challenge them.

I wouldn’t be surprised if the problem got worse rather than better too. As more people click the “thumbs up/thumbs down” button on their AI results, AI is going to increasingly use those as source materials for new results, leading to answers that are little more than robots-citing-robots that were, in turn, encouraged by people’s confirmation bias.

Hello dead Internet.

As to what the answer is for this, I can’t say, but I can theorize. I think for general AI to get better, it’s going to have to start looking to specialized AIs with deep domain expertise for answers. And in the world of science that answer, more often than not, is going to be “We’re not sure. More research is needed.” Perhaps not as satisfying as a simple “yes” or “no”, but more accurate.

Ultimately, I, like the SciPak people, am not ready to turn everything over AI now, or possibly ever. It’s fairly obvious from my simple test that good old-fashioned human intuition and decision-making will continue to reign supreme.

~#~

PS: We ran that PubMed study I referenced above through our Tessa AI Platform and she scored it solidly in overall trust. Tessa found the scientific methodology and conclusions within the study well-founded and reasoned, the authors respected, and no AI involved with the generation of the study. The only detractions Tessa found were within the publishing journal’s impact factor, and the paper’s media score, a metric that analyzes the quality of the discussion around it. Both of those metrics can change over time.

Ultimately Tessa concluded the study was trustworthy but, you guessed it, more research is needed before we can come to any definitive conclusions.

The post Are AI’s Answers Too Good to Be True? The Bias in Your Search Results first appeared on Tessa AI.

Defining trust

Before I launch into how to restore trust, I should clarify what I am talking about when I use the word. Trust, simply put, is confidence in or reliance on the ability, faithfulness, or accuracy of someone or something.

While trust is often found in the company of “truth”, the two are not the same. “Truth” is a result, whereas “trust” is focused on the means and methods used to arrive at a result. It is an important distinction because in the world of scientific research we don’t always discover a truth, but we should always be able to determine the level of faithfulness and accuracy in the means and methods used to perform that research.

In other words, it needs to be trustworthy.

Components of Trust

Trust is a sort of cognitive shortcut that allows us to rely on something without a deep analysis of that thing. If it meets certain criteria, it can be trusted. Psychologists have studied the concept extensively and identified a number of cognitive and social factors that influence how people perceive someone or something as trustworthy (and vice versa).

Even if you have not studied the concept, you are still probably familiar with many of these factors (aka “heuristics”): source credibility or authority, bandwagon effect, perceived motive and intent, consistency and coherence, willingness to engage and revise, confirmation bias, etc. In all, there are around a dozen of these elements; too many for the purposes of this article.

Instead, I have borrowed a simplified framework for trust from computer science and neural networking where they have taken the various components of trust and put them into three interrelated conceptual buckets: Transparency, Explainability, and Significance (TES).

• Transparency can be summed up as “show your work.” In scientific research it is the clarity and fidelity of the methodologies and data sources, affiliations of the researchers, funding sources, a level of detail that allows others to repeat the research described, and so on.
• Explainability is the consistency and coherence of the information and the ability to support it if it contradicts other information. It also factors in elements like the level of uncertainty, the ability to address questions and criticisms, and details that help shed light on how the conclusions were reached.
• Significance is a measure of the relative importance of both the components and results of a piece of research. Components include the type and nature of study, the reputation of the authors, institutions and publications, etc. It also factors in peer review commentary, recency, citations, and other elements that help determine the overall impact of the information.

Each component of TES has many interrelated sub-components so the weighting of each element varies based on the information being analyzed. However, placed on a scale of 0 to 100, the closer each component is to 100, the more “trustworthy” it is (though scientific skepticism makes it extremely unlikely that anything or anyone could ever have a perfect 100 across all three categories).

Restoring Trust in Science

In my previous post I identified a number of areas where the scientific research and medical communities had significant failures in each area of TES, leading to an erosion of public trust in science. Whether it was a failure to disclose financial incentives (transparency), the repeated use of fraudulent data (explainability), or the resistance of highly regarded scientists in acknowledging their inaccuracies or errors (significance), or all of them collectively, the damage done has been significant and as such, will take some time to repair.

The good news is the scientific community is more aware of these problems than ever, and many groups are now working on a variety of solutions, both traditional and technological. The quickest fixes lie in a return to the scientific principles of transparency (especially funding sources), open analysis, and a willingness to address conflicting or contrary information in a timely manner.

Longer term the solutions will most likely be technological. Open access and pre-publishing platforms have turned the research spigot wide open, doubling the amount of available research in medicine alone every two months. Parsing and processing that volume of information will require methods faster and more robust than the traditional, more measured and time-consuming methods of peer review and journal publication.

Not surprisingly, AI is playing a large role in attempting to get more control over the research deluge with numerous tech firms devising innovative and useful ways to search, process and digest information. Other organizations like RetractionWatch are doing important work in surfacing corrections and retractions in the pursuit of scientific progress.

For our part, here at Siensmetrica, we are focused on applying deep data science and multi-vector analysis to identify and score research based on TES principles. Our aim is not to supplant the rigors of the scientific method, but to provide scientists and medical professionals with an objective way to assess the trustworthiness of a piece of research—a tool for guiding further exploration. The goal is not to always be right, but to be transparent and consistent so that we can restore belief in a system which, though not always perfect, can be relied upon to be heading in the right direction.

If we are successful, “trust the science” will be the gold standard once again.

The post How to Restore Trust in Medical Research first appeared on Tessa AI.