How to Lie with AI (a cautionary blog)
2024 marks the 60th anniversary of the publication of How to Lie with Statistics, a book that exposed how easy it is to manipulate data, algorithms, and output. The anniversary presents a good opportunity to reflect on the fact that much of today’s advanced analytics, such as machine learning (ML), natural language processing (NLP), and generative artificial intelligence (AI), are based on probability theory , a fundamental part of statistics.
Public health researchers use probability theory to understand the relationship between exposure and health effects (for example, disease transmission, vaccine effectiveness, injury prevention, and much more). Advanced analytics hold the potential to improve public health in many ways, but they can also be used in ways that could harm our health. For example, some companies are exploring how to use AI to develop more effective marketing strategies for the tobacco industry .
"Now, more than ever," writes Mathematica's Amanda K. , "it’s critical that public health officials—or anyone looking for real answers—continue to make three principles the bedrock of their work."
Market Intelligence I Evaluation Specialist I Consumer Insights I Social Researcher
2 个月Totally agree with this summation. I made a recent write-up about this on Medium stating the need for decision makers to pay close attention to the research process as a way of validating the quality of data feed into AI, building more confidence on the results/outputs churned out. The community stakeholder lens sound also incredible from a public health persepective.
?? Policy to practice through digital modernization | ?? Regulation and compliance | ?? Law student
2 个月Great post. The LEEP that Mathematica has formed sits squarely within the public health framework of community-based participatory research. Using the community stakeholder lens to keep algorithms in check and honest requires a supportive plain language “interpreter,” so to speak, whom can also share back the impacts the communities have made on the output of their contribution. In other words, as responsible researchers, program managers, and agencies, we cannot forget that we have to give back outcomes information to communities - not just take design input. As I’m not an NLP/ML/AI expert at the command line, this question may be totally ignorant: can we introduce opportunities for transparency and reproducibility in this era of AI that will show proof to the equity of our algorithms? I know the expense and customization to continue fueling an appropriate model becomes the most proprietary part of the work at times, but is there a way to share that doesn’t give up the value entirely? If so, as public health practitioners, are we doing that now? Are we doing it enough?