Interactive Evolution Instead of Default Parameters

Daniel Ashlock, Cameron McGuinness, and Joeseph O'Neil
Submitted to CIBCB 2015

Abstract PDF eprint

Tools for processing biological data often have many parameters, but most users simply use the default settings. Such software often has a large number of controls or user specified parameters. This means that there can be problems with teaching users to use even standard bioinformatic tools effectively. This study prototypes a technique called a show- me-more interface that uses human-in-the-loop evolution to permit an untutored user to operate a complex software tool that designs images of flowers. This task is intended to permit research on managing complex parameters for users that do not understand them without the added complexity of working with biological data. Users are given two specific and two non- specific tasks and the results of their design efforts are displayed and discussed. The basic concept of show-me-more control has broad applications for permitting casual users to manipulate complex tools in a simple and transparent manner. Careful design can minimize the number of clicks needed for a user to reanalyze data, reducing the potential for user fatigue and attendant error.