Northwestern University (Jan 21): The stability-plasticity dilemma in predictive neural network models of semantic memory

​Mike Jones and Steve Lewandowsky were honored for their service to the Psychonomic Society at the recent Annual Conference. 

Mike steps down December 31 after six years as Editor-in-Chief of Behavior Research Methods (BRM), the world's top methodology journal. Mike and his team have built BRM into the highest impact journal in the 60-year history of the Psychonomic Society, and BRM now has a higher impact factor than all major competitors. 

Mike: Your watch has ended!

This year's Presidential Symposium at the Annual Meeting of the Society for Computation in Psychology (SCiP) featured the winners of the Early Career Award from the Federation of Associations of Behavioral and Brain Sciences (FABBS): Laura Allen (2019 winner), Rick Dale (2016 winner), and Mike Jones (2012 winner). 

Mike's talk was titled Limits of Deep Learning: Modeling Human Learning to Inform Machine Learning 

 Congrats to Jessica Montag on winning her first NSF grant as PI to study early word learning environments and the effects of early picture book reading. Jessica also won a prestigious J. S. McDonnell Foundation Award to develop new simulation methods to understand optimal early word learning environments, expanding on our recent paper in Cognitive Science. Jon and Jessica recently moved to University of Illinois at Urbana-Champaign, where they are setting up their new labs. 

Mike will give one of the invited State-of-the-Art lectures at this year's International Congress of Applied Psychology in Montreal. Every four years, the IAAP organizes a world congress of applied psychology which serves as a review of advances in applied psychology and unites several thousand psychologists from all over the world. 

Mike's lecture is titled Advancing Psychological Theory by Mining Big Data and is intended to be a broad survey and taxonomy of current cutting edge work being conducted to use big and naturalistic data sources and sensors to reverse engineer the psychological mechanisms that generated the behavioral traces. 

Massive congratulations to Dr. Melody Dye, who is the recipient of this year's Glushko Dissertation Award from the Cognitive Science Society. This very competitive award is highest honor in our field for a graduate student, won for Melody's doctoral dissertation "Bridging Levels of Analysis: Learning, Information Theory, and the Lexicon."   It also comes with $10k, so that isn't bad either. 

Melody will receive the award and give a talk on her dissertation work at this year's Cognitive Science Society Conference in Madison. 

Congrats to lab alumna Melody Dye, who was presented with the Outstanding Dissertation Award from the IU Cognitive Science Program for her work adapting information theory to study optimal language and nominal systems design. 

Melody is currently a Sloan/Moore Data Science Postdoctoral Fellow at the School of Information, UC Berkeley. 

The lab's IES funded project with the Karpicke lab at Purdue was briefly covered in the recent issue of Science. The special piece was  spotlighting recent projects funded by IES to translate science of learning findings from the lab into practice in the classroom to boost learning and retention. 

READ FULL ARTICLE HERE. 

Farewell to Dr. Fatemeh Torabi Asr! 

Fatemeh has been the leader of the IU end of our IES project on automated learning technologies with the Karpicke Lab at Purdue Unviersity. Fatemeh leaves us to take a position at Simon Fraser University in Canada. 

Best of luck, and you will be greatly missed!

Huge congratulations to Dr. Melody Dye on successfully defending her PhD dissertation!

Melody has taken a position as a prestigious Berkeley Data Science Fellow, working with Marti Hearst at the UC Berkeley Institute for Data Science.

Farewell Mel, and good luck from us all!

Huge congratulations to Melody Dye for winning this year's David Marr Award from the Cognitive Science Society! The Marr Award is widely considered the most prestigious CogSci student award in the world...only one is given out each year for the best student paper. The award will be presented to Melody at the 2017 CogSci conference in London, UK. 

Melody won a very competitive position as a Berkeley Data Science Fellow, and will begin working with Marti Hearst at the UC Berkeley Institute for Data Science as soon as she defends her PhD next month! 

Congrats Mel!

Yogi's Bar and Grill has been the standard Friday afternoon watering hole for the cognitive science labs for well over a decade now. Sadly, it abuptly closed down, and we have no good home for after lab meetings. 

Many great ideas came from meetings in the area pictured to the left. Here is the final ever picture of the group brainstorming at Yogi's. 

Congrats to Dr. Janelle Szary, who leaves the lab next month to take up a prestigious Insight Data Science Fellowship in New York. Janelle will be working with the best in the world to bridge her computational skills from academia into industry to solve real problems! 

Best of luck from us all Janelle! 

Mike will present Modeling Human Learning to Inform Machine Learning as one of the invited plenary speakers at this year's High Performance Computing Symposium, Canada’s premiere Advanced Research Computing (ARC) conference, bringing together top researchers from across Canada and around the world, as well as major industry partners.

IU Bloomington Emerging Areas of Research program makes inaugural grant


FOR IMMEDIATE RELEASE
BLOOMINGTON, Ind. -- Can machines learn to think like children? An interdisciplinary team of cognitive scientists, neuroscientists and computer scientists at Indiana University Bloomington has received the campus’s inaugural Emerging Areas of Research funding award to explore that question.
Called “Learning: Brains, Machines and Children,” the first Emerging Areas of Research initiative is led by Linda Smith, Distinguished Professor and Chancellor’s Professor of psychological and brain sciences in the IU Bloomington College of Arts and Sciences. The research team will receive up to $3 million for the four-year project.  

[News Release]   [Project Description]

Lab Alumus Dr. Brendan Johns, now Assistant Professor at University at Buffalo presented a keynote talk at this year's Society for Computers in Psychology, in Boston. Brendan was invited as part of this year's focus topic on Cognitive Technologies. 

The title of Brendan's address was ​Content-Driven Machine Learning: Using Lexical Variability to Optimize Models of Natural Language, and this was the first introduction to the field of our new methodology for Experiential Optimization. 

We have released a github page containing a python implementation of gc-lda for identifying functional regions in the brain:

https://github.com/timothyrubin/python_gclda

Paper,    Supplemental materials,   ​Poster

Congrats to Dr. Tim Rubin, who is leaving his position in the lab to join the NLP Data Science Team at SurveyMonkey in Palo Alto. 

Tim ran our lab's arm of the multisite NeuroSynth project for 4 years, and created the NeuroTopics approach to co-discovery of functional-anatomical brain regions and cognitive processes from the neuroimaging literature. 

http://www.theonion.com/article/study-us-wastes-2-million-hours-annually-figuring--50764

But think of the massive impact on the economy if we could find an optimum solution strategy!

Big Data in Cognitive Science is in production, and should be on shelves in the fall. If you're super eager, you can always pre-order a copy on Amazon. 

Read Introductory Chapter: Developing Cognitive Theory by Mining Large-Scale Naturalistic Data 


​Table of Contents

Congratulations to Brendan Johns and Mike Jones for winning this year's Canadian Journal of Experimental Psychology Best Article Award, co-funded by CPA and CSBBCS. The award will be presented at both the 2016 CPA Convention and the 2016 CSBBCS Meeting.

Johns, B. T., & Jones, M. N. (2015). Generating structure from experience: A retrieval-based model of language processing. Canadian Journal of Experimental Psychology , 69, 233-251. ​

The lab welcomes Dr. Janelle Szary as a new postdoctoral research associate. Janelle recently completed her PhD at UC Merced working with Rick Dale and Chris Kello. She specializes in complexity theory and understanding information flow in complex networks. Janelle will be extending her work on collaborative cognition while at IU, studying how dyads and groups learn and solve problems together. 

Our new study, just published in Psychological Science, is getting a bit of press coverage now. The general finding was that interleaving ambient speech with reading picture books to young children introduces precious statistical variability that can potentially explain developmental phenomena as a function of the statistical environment (i.e., independent of maturation, per se). Read the press summary at Pacific Standard. Original paper: 

Montag, J. L., Jones, M. N., & Smith, L. B. (2015). The words children hear: Picture books and the statistics for language learning. Psychological Science.

Update: OK, now our study has been mentioned in the New York Times! It is fantastic that the general public is taking notice of the importance of statistical diversity in early language learning, but most of the press coverage kind of skims over the important finding. It is true that children experience more unique word types when we read from books vs. the ambient speech environment--I mean, how could that not be true? Reading Dr. Seuss makes me say lots more weird words than I do when I'm cooking. The important point is that the diversity distribution has a very different shape in books vs. speech. Now this makes sense. When humans were able to offload memory to external symbolic devices (such as books), we loaded them with connections in the tail of the Zipf Distribution. When we read to kids, they don't need to know what the words mean. It introduces them to statistical connections that boost everything else they already know. This may well be the reason that we see many so-called developmental phenomena such as the "vocabulary burst"

This makes the point clearer with an artificial language experiment: 

Jones, M. N., Johns, B. T., & Recchia, G. (2012). The role of semantic diversity in lexical organization.Canadian Journal of Experimental Psychology, 66, 121-132.