Unity implementation of Daw's Four-Armed Restless Bandit; a reinforcement learning task used recently in a study assessing value-based decision-making in Parkinson’s disease apathy (see here for summary.
Intending to be a portable; OS/ device agnostic task. Builds work for for webGL, iOS, Mac & Windows.
Original motivation was to deliver task on MR compatible tablet (iPad- see MRiOS branch). See WebGL build for a demo/ to play through.
To build; download Unity (see dependencies); and import & build C# code should be accessible and understandable; and as such doesn't fully leverage advanced features of the C# language. Task in Main matches Daw's Restless Bandit params (see the supplementary material for details); but these are exposed in the inspector to allow anyone to play about with them.
Unity 2022.3.16f1; 2D (URP) (Packages in package list currently not clean and include packages from other projects)
Follow Unity Build Processes but just ask if help needed; always happy.
Data read out to the Unity Application Persistent Data Path; WebGl results are written onto the browser console. Looking to improve this in a low tech way (eg just read out something sensible to idbfs).
These branches have scripts that should email the data at the end of the task to the experimenter. (The timing and play parameters may be altered in these branches compared to original).
To do this with minimal effort, open a new gmail account and add 2-factor auth. Once done, SIGN IN TO THE NEW ACCOUNT then head to https://myaccount.google.com/apppasswords where you can create an app password. The Email Data (script) component attached to GameManager has empty fields in the inspector- fill with this new email and this 'app password' (not the gmail account password).
For code/ build @i-brnrd on here or via University of Dundee. For use of it try here instead.
Isla Barnard (developer)
Graham Mackenzie, Will Gilmour, Tom Gilbertson (lead)
To Nathanial Daw for Daw's Four-Armed Restless Bandit
See here.