Interview Outline
This interview, hosted by Judith Hurwitz, chief evangelist at Geminos, features a discussion with Paul Hünermund, co-founder of causalscience.org, and a leading expert in the field of causal inference and artificial intelligence. The conversation highlights the increasing relevance and adoption of causal AI in industry, largely driven by a dedicated group of specialists within organizations. There's a growing realization that many of the questions faced by data scientists have a causal component. However, there is still a mismatch between the questions being asked and the predominantly correlation-based tools being utilized. This mismatch often causes frustration among data scientists and clients alike.
The conversation also debunks the myth that having more data will solve all problems. Instead, Paul advocates for asking the right questions and choosing the appropriate tools to address those questions. He mentions some key use cases for causal AI, such as providing explainability and interpretability for AI systems. He further emphasizes that these systems increase trust and decision-making authority for algorithms.
One practical example discussed is the use of causal AI in A/B testing, particularly in the e-commerce and online world. The overwhelming number of hypotheses that could be tested can often become overwhelming, and Paul suggests that causal AI can efficiently build on past experimental data. This can be particularly beneficial in unprecedented situations like the Covid-19 pandemic, where past experimental knowledge might still be utilized productively.
However, to conduct these studies successfully, one must bring in not only data but also domain knowledge. This could come from experts in the field or be data-augmented using causal discovery techniques. Paul asserts that though this might sound intimidating, simple qualitative assumptions about differences can be sufficient to create causal models, and the rest can be supported with data.
Regarding the future milestones of causal AI, Paul sees the need for more tools and educational materials. He also emphasizes the importance of bringing together people from different scientific disciplines and industries to facilitate conversation and collaboration. This, he believes, will greatly benefit the field of causal AI.