A talk to Hausi Müller's group at UVic

I gave a short talk to Hausi Müller's group here at UVic about my research plans. I've posted the slides so you can take a look. Note the slightly provocative title.

Why Do Computational Scientists Trust Their Software?

View more Keynote presentations from jpipitone.

Some of the feedback I got from the group:

• You could try looking at how defect rate changes over time. Use the trends as comparison points, or build a prediction model and then compare parameters of the model. (see ISSTA)
  
• Can you really investigate code quality separately from the concerns of model quality?
• How would you (well, the climate scientists) know a perfectly correct piece of software if they had one?
• The criticality of a defect has many dimensions.
• Consider code churn in estimation of code quality (high churn with low defects -> high quality process?). Or simply just making the churn data available might help ground the defect density data.

I actually enjoyed giving this presentation. Partly because it was to a small group and I'm much happier (read: less nervous) in discussions rather than presentations (I have been known to get incredibly nervous in front of groups). More than anything though, it was helpful to put together the slides as a way to articulate what I'm doing. It's the first time I've felt that I have something resembling a cohesive story to tell.