New Frontiers at the Intersection of Artificial Intelligence, Medicine and Public Health from Optimization to Oral Communication

Drug overdose deaths resulting from substance use disorder (SUD) have been on the rise. For patients experiencing SUD, long-term and multi-disciplinary care is vital as this population is at high risk of recurrence. With the increasing incidence of SUD, clinical care teams are at risk of burnout. Use of a chatbot is proposed to alleviate pressure to the healthcare system, as it has the potential to be scalable and cost-effective. This study is the product of a collaboration between Dimagi, a corporation developing this chatbot, UCLA, the academic partner providing research support and chatbot evaluation, and Massachusetts General Hospital, the clinical partner who will deploy this chatbot with their patients. Four beta-testers were used to examine feasibility of implementation, and we will be comparing substance use at baseline vs. at the end of the study for 55 patients who will be utilizing this chatbot. For this presentation, I will discuss the pros and cons of using a chatbot and how we design a study to train and evaluate large language models where special care must be taken when tuning the chatbot's sensitivity to matters that should be escalated to a health professional, as stakes can be much higher in a healthcare setting as compared to chatbots in service delivery sectors. 

Nature-inspired metaheuristic algorithms are important components of artificial intelligence, and are increasingly used across disciplines to tackle various types of challenging optimization problems. This paper demonstrates the usefulness of such algorithms for solving a variety of challenging optimization problems in statistics using a nature-inspired metaheuristic algorithm called competitive swarm optimizer with mutated agents (CSO-MA). This algorithm was proposed by one of the authors and its superior performance relative to many of its competitors had been demonstrated in earlier work and again in this paper. The main goal of this paper is to show a typical nature-inspired metaheuristic algorithmi, like CSO-MA, is efficient for tackling many different types of optimization problems in statistics. Our applications are new and include finding maximum likelihood estimates of parameters in a single cell generalized trend model to study pseudotime in bioinformatics, estimating parameters in the commonly used Rasch model in education research, finding M-estimates for a Cox regression in a Markov renewal model, performing matrix completion tasks to impute missing data for a two compartment model, and selecting variables optimally in an ecology problem in China. To further demonstrate the flexibility of metaheuristics, we also find an optimal design for a car refueling experiment in the auto industry using a logistic model with multiple interacting factors. In addition, we show that metaheuristics can sometimes outperform optimization algorithms commonly used in statistics. 

Swarm intelligence (SI) algorithms simulate the emergent intelligence of flocks or swarms of animals to solve complex problems. These algorithms are particularly useful for solving complex optimization problems that are difficult with traditional methods. A major advantage of SI algorithms is that they can avoid local optima, work on black-box functions, and are easy to implement. These advantages make SI algorithms suitable for designing efficient experiments in toxicology, especially for designs with complex objectives. In this presentation, I will introduce several design problems in dose-response toxicology where SI algorithms have advantages over standard methods. The goal of the presentation will be to demonstrate the effectiveness and ease of use of SI algorithms for solving complex problems in experimental design. 

Communication between clinicians and patients is essential for healthcare delivery. As a result, medical educators devote time in curricula for communication skills training. Active practice takes place when students role-play as clinicians with standardized patients (SPs). Due to competing clinical training demands and limited resources to implement SP programs, medical researchers are developing large language model (LLM)-driven simulated patients to supplement SP programs. Student learners interact with LLM simulated patients through simulated clinical encounters, either on a computer screen or through virtual reality hardware. I present the results of a pilot study as illustration. A research team and I developed a simulation where medical students role-play as a primary care physician to discuss the results of an abnormal mammogram with a LLM simulated patient. We pilot tested the simulation with 10 medical students at the University of California, Los Angeles to evaluate its feasibility and acceptability in a medical school setting. Along with the study findings, I discuss human-centered design principles and prompt engineering strategies we used to create the simulation, and lessons learned in selecting and working with LLMs that are generalizable for other LLM applications.