| 
  • If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • Stop wasting time looking for files and revisions. Connect your Gmail, DriveDropbox, and Slack accounts and in less than 2 minutes, Dokkio will automatically organize all your file attachments. Learn more and claim your free account.

View
 

Honour's Thesis ideas

Page history last edited by Jim Davies 1 year, 8 months ago

Fourth-year undergraduates in cognitive science (and psychology) need to do an honours project. This page has some project ideas. Note that if you are a psychology student, you will need a co-supervisor in psychology if you want to work with Jim Davies.

 

Programming Projects

 

Lake: Python ACT-R and Language

Carleton has a cognitive architecture, Python ACT-R. Right now it does not have language components. This project would involve finding state-of-the-art and available language software and interfacing it with the architecture. It could be speech synthesis, natural language understanding, or natural language generation.

 

Lake: Python ACT-R and Scene Descriptions

Many graphical worlds are described not in pixels or voxels (3d pixels) but in higher-level scene descriptions. These could potentially be turned into ACT-R chunks. This project is to make software that takes one or more popular scene description languages and turns the information into information in the python ACT-R "environment."

 

Project 5.27: Replacing Sentences with Rhyming Versions

This AI would take in a sentence and outputs a version with roughly the same meaning, but with more rhymes and alliteration in it. For example, it might take in the sentence "He slaps his hands on the lamp post, but maintains that he sees an apparition" and output "he hits his fists against the post but still insists he sees a ghost." There is already a system that measures rhyme density in text: if you google "sourceforge rhymeanalyzer"

 

Vector memory and Spatial Detectors

     Ph.D. student Matthew Rutledge-Taylor is working on a vector space memory model that has some impressive properties. In this project, you will apply this model to the domain of visual memory. Undergraduate Connor Smith has been leading a project creating psychologically-realistic visuospatial detectors in software. This project involves generating an imagined scene using Smith's data and Rutledge-Taylor's memory model. Co-supervised with Dr. Robert West.

 

Procedural Generation of Descriptions

The field of procedural generation is a part of graphics as much as AI. But how do we create creative things that are described in terms of features? For example, if we have a list of monster features, can we make an AI that uses this list to create a description of a new monster? How about a culture, or a personality? This project devises an AI that randomly creates descriptions of things, but has dependencies between them (e.g., a swimming creature is more likely to have scales). 

 

Philosophical/Theoretical/Literature Review Projects

 

Review of Functionalism

     Explore the roots of the philosophical position of functionalism and what's being written about it today, especially in relation to emotions and pain in artificial and distributed cognitive systems.  

 

Project 5.4 Conceptual Analysis of Bottom-Up vs. Top Down

These terms get used in several ways in cognitive science. A short paper, perhaps for submission to the Cog Sci conference, could be written about the various ways they get used. 

 

Project 5.21 Functional and Computational descriptions of what brain areas do

It's common for some brain areas or structures to be used in a variety of cognitive tasks. This suggests that each structure is doing some kind of computation that is useful for many tasks. How can we describe these functions? Depending on how this project goes, it might be more about the idea of this, or taking a specific brain structure (e.g., the Amygdala) and trying to come up with a computational account of what it does across tasks.

 

Empirical Psychology (running participants) Projects

 

Project 5.33: Memories from Further in the Past are More Prototypical

Previous work has shown that imagining things happening in the past (real or fiction) are imagined as more prototypical than things imagined in the future. 

Gilbert, D. T., Morewedge, C. K., Risen, J. L., & Wilson, T. D.(2004). Looking forward to looking backward. Psychological Science, 15, 346-350.

But as of yet nobody has shown that imagining things further in the past or future increases the effect size. That is, if you imagine something far in the past, is it less prototypical than thinking of something recently in the past? Run an experiment to test this. 

 

Project 5.30: Can you remember colours visualized as being behind you? 

Some scientists claim that there are two kinds of picture-like imagery: visual and spatial. Visual imagery is 2d and has size and color and object identification, and spatial imagery is 3d and has only locations of things in space (ignoring things like texture and colour.) Visual imagery is thought to be analogous to your visual field, but in the head. So if you imagine something behind you (without cheating*), then it shouldn't retain very well visual properties like colour. So in this experiment participants look at colours, say a particular shade of red, and then try to hold the image of that colour in imagery, but as though it were behind them. Another group images it normally. Later, we predict that people who imaged behind would have poorer recognition of exactly what shade of colour it was. 

* cheating would be to reorient oneself in the imaginary space so that they are looking at themselves in the 3rd person, with the colour behind them, technically, but not behind the imaginary point of view.

 

Detector Testing

     Our laboratory has created AI spatial relationship detectors, such as "above" and "in between." These need to be tested psychologically. Do the AIs work the same way people do? Co-supervised with Robert West and in consultation with Shelli Feist.

 

 

 

Comments (0)

You don't have permission to comment on this page.