Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

The impact of noise on memorability in minds and machines

Martin Wiener¹; ¹George Mason University

Presenter: Martin Wiener

Recent evidence suggests that image memorability, the intrinsic likelihood that an image will be remembered, relates to the processing efficiency of the image. One unknown impact is the effect of introducing noise into this process. By comparing a variety of memorability prediction models against human performance on a memory task with differing noise levels, I found a striking difference between the impact of noise on each: noise degrades memorability in humans but enhances it in machines, but only for low memorability images. In contrast, high memorability images are resistant to noise in humans, but degraded in machines. I present here a new memorability model that incorporates inter-image differences, is robust against noise, and more closely matches human performance.

Topic Area: Memory, Learning & Knowledge Structures