What is post theory science?

Thank you for tuning in to this audio only podcast presentation. This is week 76 of The Lindahl Letter publication. A new edition arrives every Friday. This week the topic under consideration for The Lindahl Letter is, “What is post theory science?”

Last week some difficult questions were asked about overcrowding within the field of machine learning and the effect of that on engineering colleges [0]. This week things are getting even deeper into the hard philosophical questions we are starting to face. Working to answer the question, “What is post theory science?” will require a lot of consideration and a good bit of digging around. A lot of the ML things that are bubbling up right now in terms of ethical conundrums and rapidly changing ML delivery use cases are just ahead of the next wave of publications and methodology research articles, manuscripts, and textbook updates. This is another topic that I think is going to end up getting a lot of coverage at some point. Just like the potential effects of overcrowding within the machine learning space limiting other research within engineering colleges. The compounding problem from that is that the overcrowded research happens to be very shallow within the ML space resulting in papers without real and lasting contributions to the field. 

Let’s focus on the question at hand, “What is post theory science?” This is an interesting scenario to have in existence. Machine learning models can be built to seek out solutions without any theoretical methodology being applied to the potential solution. That type of possibility created a situation where Laura Spinney of The Guardian asked the question, “Are we witnessing the dawn of post-theory science?” [1]. Within that analysis Laura called back to an article from 2008 by Chris Andeson in Wired magazine titled, “The End of Theory: The Data Deluge Makes the Scientific Method Obsolete” [2]. In this scenario without using any theoretical basis to create a testable hypothesis or any classic research method a machine learning use case could derive defendable answers. You could do a similar task with big data or just data in general. Things can be observed and built into a postulate or observation that is not directly based on scientific theory. We can learn things that are seen as being objectively true, but are not derived from theory. In this case that type of effort could very well be called a post theory scientific research methodology. You could build an AI model that aims to understand elements of the universe and it could just be allowed to run and work on that effort. Within that proposition the AI model could test and work with model settings that yield results, but are not based on theory. It could just be using random configurations or working down a path that was independently derived. 

A search within Google for “post theory science” does not yield a ton of results. At the time this post was written only about 1,740 results existed within the knowledge graph. I’m truly curious when research method books will contain post theory methods. For post theory science to really pick up steam as a method of research in academic institutions it will need to start showing up in research methods textbooks. I have a couple of them on the bookshelf next to me and while they cover mixed methods and a variety of approaches, nothing within that very large book presumes the idea of post theory science. I’m going to guess at some point that will change here in the next couple of years. We are going to see findings and research coming out of the ML space at record levels and a portion of it will be delivering results, conclusions, and data that are not derived from the traditional scientific method. 

Links and thoughts:

I watched this entire video, “This Beat Up, Non-Running Omega Seamaster Has Big Potential! Vintage Watch Restoration”

“Babbage: Artificial intelligence enters its industrial age”


“What unions could mean for Apple with Zoe Schiffer”


I’m in and out on Lex’s content, but sometimes I just enjoy listening to a good conversation. “Jonathan Haidt: The Case Against Social Media | Lex Fridman Podcast #291”

“The Lab is a Disaster – WAN Show June 3, 2022”

Top 5 Tweets of the week:


[0] https://nelslindahl.substack.com/ “Is ML destroying engineering colleges?”

[1] https://www.theguardian.com/technology/2022/jan/09/are-we-witnessing-the-dawn-of-post-theory-science 

[2] https://www.wired.com/2008/06/pb-theory/ 

What’s next for The Lindahl Letter?

  • Week 77: What is GPT-NeoX-20B?
  • Week 78: A history of machine learning acquisitions
  • Week 79: Bayesian optimization
  • Week 80: Deep learning
  • Week 81: Classic ML algorithms

I’ll try to keep the what’s next list forward looking with at least five weeks of posts in planning or review. If you enjoyed this content, then please take a moment and share it with a friend. Thank you and enjoy the day!

Leave a Reply