The problem of influencing preference toward products on social networks has attracted considerable attention over the past couple of decades. Previous approaches have suffered from two subtle yet significant drawbacks. The first is that they model consumer decision making as best-response, deterministic maximization of some numerical utility. The second is that their decomposition of utility does not include influence by marketers for the respective companies.
Matthew Reyes casts consumer decision making within the framework of random utility. Random utility theory views so-called utility as a parametrization of observed frequencies of choice. The decomposition of utility corresponds to variables that are either observable through data collection or under the control of an external agent, in this case a company. The decomposition of utility that Matthew presents explicitly includes influence by marketers from two competing companies.
Incorporating the marketer into the model of consumer decision making allows a company to evaluate the effect of different marketing allocations on the evolution of preferences on the network. The combination of a random choice model and the inclusion of marketers into the model allow this important problem to be cast in the reinforcement learning paradigm. Matthew outlines a simplified scenario illustrating the steps in a company’s allocation decision, from learning parameters from data to evaluating the consequences of different marketing allocations.
Matthew Reyes is a consultant and an independent researcher developing a reinforcement learning-based approach to influence the maximization on social networks. Previously, he spent more than four years at MIT’s Lincoln Laboratory. He holds a PhD and MS in electrical engineering systems from the University of Michigan, Ann Arbor, and an MS and BS in math from Wichita State University.
Matthew Reyes is a contractor at Technergetics working on deep learning and FPGAs. He earned his B.S. and M.S. in mathematics at Wichita State University in Wichita, KS. He then earned his M.S. and Ph.D. in EE:Systems at the University of Michigan in Ann Arbor, MI doing a thesis on compression of Markov random fields. He worked for four years at MIT Lincoln Laboratory doing work on sensor calibration. From early 2015 to early 2019, Matt conducted independent research on compression, belief propagation, and interpolation of Markov fields, and is currently developing a model of social decision-making based on random utility.
Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?
Join the conversation here (requires login)
©2019, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • firstname.lastname@example.org