Our world is full of a dynamic array of sights and sounds. The brain must adapt to this complex environment by prioritizing some signals over others. One way the brain does this is by constantly updating the influence of various decision outcomes on existing beliefs. These influences on adaptive learning can be decomposed into distinct factors and modeled using computational modeling. We have designed and implemented a spatial prediction task that is disguised as a helicopter game. In this task, the user will move a basket or bucket to collect goods that are dropped from a helicopter. Various aspects of the gaming environment change, including the location and visability of the helicopter, as well as the value of the goods dropped. This task allows us to measure learning in response to stimuli and to examine the influence of three distinct components on learning, including, (1) surprise– or novelty- driven belief updating, (2) uncertainty-driven belief updating, and (3) reward-driven belief updating. Previous work has shown that participants are rational in updating their predictions based on uncertainty-driven belief updating. Our current work suggests that individuals with autism show impaired performance on this task, as measured by increased reliance on surprise-driven updating, due to a bias to overreact to novelty.
We are a team of researchers interested in learning more about children and adults with and without brain disorders. Generally, we are interested in designing applications that are engaging to all ages, but allow us to capture meaningful experimental data that will help us understand how learning is different in children with developmental disorders like autism. Ultimately, we hope this will lead to more refined medical diagnoses.
We hope to develop a more engaging version of this helicopter game by adapting this task using web-based technology. It is currently programmed in Matlab and requires a few locally-built Matlab toolboxes. By adapting this task using a web interface, this will be usable on a larger scale (patients can access from home) and allow us to collect more robust datasets without requiring users to visit the clinic.
- Creation of a web-based version of the helicopter game.
- The application should store trial-by-trial data on each user, including information related to what type of goods are dropped, location of the helicopter and bucket, and other variables, such as whether the helicopter is visible or not.
- Overall presentation of the game should be improved with more engaging graphics.
- Automatic analysis/extraction of relevant variables and visualization of these variables is a secondary goal (but at least raw data should be extracted).
This project has the potential to change how we are able to conduct research. Research is increasingly using web-based technology in order to acquire large data sets. This tool has the potential to be used nationally by many researchers interested in developmental disorders and learning, more generally.
Must be able to store information in a way that can be analyzed later
- We can provide expertise and code from current Matlab version of the game.
- We can provide expertise on the variables that are manipulated within the game.
- We can provide children and adults to help test the tool’s user-friendliness.
Point of Contact
Dr. Vanessa Troiani
Geisinger Autism & Developmental Medicine Institute