Event Details

Keith Burghardt’s Title (Ashwin Shreyas Mohan Rao will be co-presenting)

“Humans are not Robots: Rules of Thumb, AI, and the Pervasiveness of Polarization”

Keith Burghardt

Heuristics are pervasive in our day-to-day life. These “rules of thumb” such as assuming popularity is synonymous with importance (e.g., the Bandwagon effect), reduce cognitive load but can also nudge people to unexpected and suboptimal outcomes. In this talk, I will show examples from my own research of how these simple rules of thumb show up repeatedly, and can negate many benefits otherwise apparent in crowdsourcing and recommendation systems. First, I will show that heuristics make finding quality content on crowdsourcing sites more difficult, but this can be mitigated with a simple algorithm. Next, I will show evidence that heuristic-based data, when fed into recommendation systems, can enhance polarization and allow for hate groups and misinformation to proliferate. Finally, I will conclude with suggestions on how correcting for these biases may make better AI systems.

Fred Morstatter’s Title (Daniel Benjamin, Bahareh Harandizadeh and Yuzhong Huang will be co-presenting)

“The Benefits of Hybridization: From Forecasting to Cultural Modeling”

Fred Morstatter

Humans are better at reasoning about qualitative data, while machines can process quantitative data at a scale beyond human capacity. Fusing the two sources of knowledge offers the ability to better understand the world around us, and to make more accurate predictions about the future. In this talk, we will discuss a series of approaches to hybridization. First we will talk about hybridization as it applies to forecasting, presenting a novel approach to aggregation that identifies the best human and machine input for a given problem. Then, we will discuss how prediction markets can be hybridized to yield more accurate predictions. Finally, we will discuss how human knowledge can be used to better understand and represent cultural values.

Speaker Bio

Keith Burghardt

Burghardt is a computer scientist at the USC Information Sciences Institute who specializes in understanding human behavior with physics-inspired models. Burghardt received multiple awards including a 2015 Conference on Complex Systems Starred Paper award and a 2016 ISI Director’s Intern Award. Burghardt organizes the USC/ISI AI Seminar series, and has organized the S3B2-ML Summer School and CKIDS student research projects. Burghardt received a BS in Physics (Magna Cum Laude with High Honors) and PhD in Physics at the University of Maryland in 2012 and 2016, respectively.

Fred Morstatter

Fred Morstatter is a Research Assistant Professor of Computer Science at USC and a Research Lead at ISI.  His research focuses on understanding biases that occur in social data, cultural modeling, and developing hybridized AI systems. He has been a key contributor to ISI's effort under IARPA's HFC program and is the PI of VENICE, a hybridized system to uncover implicit cultural knowledge. Contact him at [email protected]