Computer Science Program and Dean of the College Present
Improving the Impact of AI with Visual Affordances
Tuesday, November 29, 2022
RKC 111
5:00 pm – 6:00 pm EST/GMT-5
5:00 pm – 6:00 pm EST/GMT-5
Dylan Cashman, Senior Expert, Data Science and Visual Analytics, Novartis
Artificial Intelligence models are used in many critical applications. They both enable new technology such as robotics and self-driving carsas well as replace existing decision support in human-facing use cases like chat bots or loan assessments. These critical applications hold high risk if the AI model is biased, inaccurate, or not trusted by its intended consumers or those affected by its decisions. In this talk, several methods for improving the impact of AI are presented by making use of visual affordances and human-centered design. First, a black box model selection tool is presented that enabled subject matter experts to beat an automated machine learning algorithm. Second, a visual analytics tool for human-guided neural architecture search demonstrates how users can shape an AI learning process to fit their use case. Lastly, ongoing work into human-centered design of visualizations for explaining models is presented. Dylan Cashman is a Senior Expert in Data Science and Advanced Visual Analytics at Novartis Pharmaceuticals in their Data and Artifical Intelligence division. He received his Bachelors of Science in Mathematics at Brown University, and his Masters and PhD in Computer Science at Tufts University under Dr. Remco Chang. He has previously worked with Northeastern University, the MIT IBM Watson AI Lab, PARC, and MIT Lincoln Laboratory, and has served as a technical consultant to startups in the Boston area. He has won best paper awards at the Eurographics Visualization Conference, the Visualization Data Science Symposium, and the Workshop on Visual Analytics for Deep Learning. His research leverages visual affordances and human-centered design to improve the impact of AI across many use cases and data types.
For more information, call 845-758-6822, or e-mail [email protected].
Time: 5:00 pm – 6:00 pm EST/GMT-5
Location: RKC 111