Quantifying creativity in AI-generated podcasts

Creativity and variety in teaching methods are proven factors that enhance problem-solving, critical thinking, and understanding of the material. AI-generated podcasts are emerging as a personalized learning tool as they make learning more engaging and memorable. Enhancing the creativity of AI-generated podcasts can take personalized learning enabled by these podcasts to next level. A comprehensive and robust metric to assess creativity in AI-generated podcasts can enable personalized learning at scale. We propose a creativity score that quantifies the creativity in an AI-generated podcast using 3 important aspects of creativity: divergent thinking, emotional diversity, and analogies. We build on and use the DSI (Divergent Semantic Integration) score, a context-based score that measures the amount of divergent thinking in the AI-generated podcast. To measure the degree of dialectical emotion (or diversity in emotion) in the AI-generated podcast, we have built an emotion detector for podcast analysis. We have also built an analogy detector that detects analogies in AI-generated podcasts. Using these, we have built a system that can automatically quantify the creativity in any AI-generated podcast. Our experiments show that our score is able to detect and effectively quantify increases in creativity between baseline and tuned (creative) versions of podcasts at an 85.19% accuracy. Our creativity score and the accompanying system can be used to automatically generate, evaluate, and select more personalized and highly creative AI-generated podcasts, thereby enabling AI systems to realize the potential to dramatically improve the quality of learning around the world.