Optimizing Communication To Optimize Behavior: Towards Achieving Third Level of Shannon's Theory of Communication
COM3 Level 2
MR25, COM3 02-70
Abstract:
Shannon's groundbreaking introduction of entropy in his seminal paper on communication, nearly 80 years ago, has left an indelible mark on various fields, ranging from the inception of the internet to the development of machine learning-based large-language models. The concepts he presented in his paper have found wide-ranging applications. In his work, Shannon proposed the study of communication at three distinct levels: technical, semantic, and effectiveness. While the technical and semantic levels focus on the accuracy and meaning of the message, the effectiveness problem delves into influencing the behavior of the message receiver. These levels form a hierarchical structure, with solutions to the effectiveness problem reliant upon solving the corresponding technical and semantic challenges.
Recent advancements in LLMs have significantly advanced our progress toward solving the second level (semantic understanding). However, the third level, pertaining to effectiveness, largely remains an unsolved puzzle. In an upcoming talk, Yaman will discuss strategies to tackle the effectiveness problem. He will delve into analyzing a communicator's message, explore various messaging strategies, and detail the complete information flow from the communicator to the receiver. This includes optimizing messages to evoke specific receiver behaviors, crafting optimal images for desired receiver responses, identifying messages that produce enduring receiver effects such as memory retention, and investigating signals within receiver behavior that may enhance natural language understanding by providing insights into the underlying message.
Bio:
Yaman K Singla is a research scientist in the Media & Data Science Research (MDSR) lab at Adobe and a Google PhD fellow in the Department of Computer Science Engineering at the State University of New York at Buffalo and IIITD. His research tries to connect Content, Consumers, and Interactions. Content (e.g., a document or an image) comes to life only when consumers start interacting (e.g., reading, writing about, liking, sharing, or any kind of action) with it. Without interactions, content has no meaning. Contrary to existing machine learning technologies that primarily seek information about content (such as sentiment analysis and keyword extraction) and attempt to "learn" solely from the content itself, Singla's research encompasses the entire communication loop. This loop involves understanding the message creator, the message itself, the recipient, the recipient's actions in response to the message, and subsequent changes in beliefs or perspectives. His research carries significant implications across marketing and advertising industries and has earned recognition through publications in esteemed venues like AAAI, EMNLP, ECIR, EACL, among others. Singla has received notable awards in both industrial and academic spheres, including the Adobe Outstanding Young Engineer Award and the Indian Prime Minister's PhD Fellowship Award.