Cataloging a year of blogging
January 14, 2016 2 Comments
Happy Old New Year.
January 2016 is the the start of the 6th calendar year and the 41st month with updates to TheEGG. The reason for the large discrepancy between these two numbers is occasional months without activity. The past year was exceptional in this regard with the longest single silence on the blog between April 4th and October 26th. This means that the year saw only 29 new entries, 2 indexes cataloging 2014, a report on the EGT reading group, and an update on readership. This post is meant to organize the last year of activity for future reference, and to try to uncover common themes.
If you like lists and TL;DRs then this is for you.
Reading and a philosophical education
I’d like to say I had a good excuse for the nearly 7 month gap on the blog, but that isn’t the case. I wasn’t completely idle during that time, however. I recently started a tradition of reading a new (non-fiction) book each month. I greatly exceeded that goal in 2015. This increase resulted in a number of reflections on reading and philosophical education.
- A year in books: philosophy, psychology, and political economy (January 14)
- Passive vs. active reading and personalization (October 26)
- Emotional contagion and rational argument in philosophical texts (November 5)
- Radicalization, expertise, and skepticism among doctors & engineers (November 22)
Last year started with comments on the twelve books that I read in 2014. The most notable of which were probably Metaphors we live by, Against Method, Thinking, Fast and Slow, and The Half-Life of Facts.
Fittingly, my first posts back after the silent gap were responses to Roger Schank’s “Reading is No Way to Learn”. I discussed how mediums and our engagement with them differ on the passive-active and personalization continuums. More active engagement lets us take more away from the object, but that is not necessarily the case for a more personalized object of engagement. Similarly, the relationship between emotional and rational arguments in philosophical teaching can be murky. Emotional engagement can be very important for putting moral lessons into practice, but can also be co-opted to make us less able to critically consider those lessons.
Fostering critical and skeptical engagement with ideas is essential for avoiding radicalization and dogma. Unfortunately, education in general is not a sufficient buffer. In fact, certain kinds of education — such as those of doctors or engineers — might even make us more prone to dogma. Or at least make one more prominent among the dogmatic. Can this be avoided by reading the right kind of books or by a certain kind of (early) philosophical education? Or should we just replace our doctors by pigeons?
These 4 posts had a corpus of around 9.4 thousand words and garnered around 1.6 thousand views.
Limits of tools and pairing with problems
As a theorist, it is important for me to remain critical of my own tools and techniques. And to reflect carefully on how I select tools and problems. When does a technique provide me with new insights and when am I using it as a rhetorical fog or mathtimidation?
- What makes a discipline ‘mathematical’? (January 19)
- Truthiness of irrelevant detail in explanations from neuroscience to mathematical models (January 20)
- Five motivations for theoretical computer science (March 28) by Abel Molina
- Space and stochasticity in evolutionary games (January 28)
- Evolutionary game theory without interactions (February 13)
- Operationalizing replicator dynamics and partitioning fitness functions (February 25)
- Operationalizing the local environment for replicator dynamics (March 31)
- Abusing numbers and the importance of type checking (November 8)
- Pairing tools and problems (March 16)
At times, I feel like ‘mathematical’ is used as a statement of authority. Sometimes well deserved. Sometimes not so much. When is it used properly? And when is mathematical detail irrelevant to an explanation? I was inspired to explore the first question with a dialogue and the second by analogy to brain imagining and physicalist accounts in the cognitive sciences.
Given the prior reservations, why study a highly mathematical discipline like theoretical computer science? Because it’s fun. But if that is not enough for you then Abel provided five more families of justification from: technology, mathematics, science, society, and philosophy. All places where the tools and techniques of cstheory can be helpful.
Getting more concrete, I looked at the specific tools of evolutionary game theory. First, I noting the weaknesses of (inviscid) replicator dynamics for reductionist storytelling. If you want to explicitly discuss spatial or stochastic effects, the this is not the tool for you. Worse yet, you can get dynamics of any game even without direct interactions between the agents — the things that is suppose to define a game — just through the consumption of a common growth medium. Thus, even if you system follows the curves of EGT, it is not enough to conclude a reductionist account. For this reason, and to convince myself that EGT is not irrelevant, I shifted to a phenomenological perspective — cutting out the often faulty reductionist middle man between theorist and experimentalist.
As you leave the comfort of reductionism, you lose some intuitive and appealing features like physical units and the implicit measurements associated with them. In an operationalist setting, you have rethink the types of your numbers and observations, and develop a logic of measurement from scratch. You then have to actually adhere to that logic, and make sure your calculations type-check.
All this second-guessing of tools might give the impression that I’ve been trying to find the best tool for the job. But that impression would be mistaken. Knowing the limits and powers of your tools allows you to adjust how you pick problems because what matters in the end is how you pick your tool-problem pair. To do innovative work, you often have to deform both the tool and the problem you are applying it to. My personal measure of success for a project is one where you advanced both the tool and insights into a problem. I give a historic example of this approach in action with Turing’s solution to the Entscheidungsproblem.
These 9 posts had a corpus of around 19.3 thousand words and garnered around 8.7 thousand views.
Mathematical medicine and cancer
One of the places I’ve been searching most vigorously for new problems is in medicine. Especially cancer biology. This has involved a lot of collaborations: all the posts in this section involve guests or collaborators.
- Double public goods games and acid-mediated tumor invasion (January 22)
- Pairwise games as a special case of public goods (February 21)
- From linear to nonlinear payoffs in the double public goods game (November 25)
- Seeing edge effects in tumour histology (February 4)
- Evolutionary non-commutativity suggests novel treatment strategies (Februart 14) by Dan Nichol
- Cancer, bad luck, and a pair of paradoxes (April 4) by Rob Noble
- Cytokine storms during CAR T-cell therapy for lymphoblastic leukemia (November 19)
- Evolutionary dynamics of cancer in the bone (December 6)
Together with Robert Vander Velde, David Basanta, and Jacob Scott, we found such a problem in the interaction of acidity and vascularization during acid-mediated tumor invasion. Although our resulting modeling is of the heuristic sort that I’ve grown comfortable with, I did get to explore extensions to my tools. Instead of modeling as pairwise games, we look at acidity as a public good for the cancer cells, and vascularization as a club good for the aerobic cancer cells.
Part of finding good problems, is learning to speak the language. In the case of my edge effect project with Jacob and David, this means reading histologies.
For Dan, the pairing of tool and problem came with random walks on static fitness landscapes and drug resistance. He also shared the series of (random) events that lead him to pursuing his DPhil at Oxford.
Rob found a problem in the view of cancer as a chance occurence, and how this varies by tissue type. In the process he discussed two paradoxes: Peto’s paradox in cancer — the problem domain — and Simpson’s paradox in statistics — the tool domain.
As part of leaving my comfort zone of heuristic models, I moved towards abductions by starting a project with Marco Davila on fitting individualized models of immune activity to a rich source of clinical data. This came from interactions during this year’s IMO workshop and used ideas we developed at the 2013 IMO Workshop.
Finally, Pranav Warman, David and I worked on simple EGT models of the dynamics of prostate cancer that had metastasized to the bone. This last post of the year, was also a nice reminder of some of the first papers that got me interested in the use of EGT in cancer a few years ago.
These 8 posts had a corpus of around 15.6 thousand words and garnered around 7.6 thousand views.
Evolution of social brain and behavior
TheEGG also returned to old problems in the social sciences. Using evolution to inform our understanding of social behaviors like our ethical codes, learning, false beliefs, and ethnocentrism.
- An approach towards ethics: primate sociality (January 24) by Alexander Yartsev
- An approach towards ethics: neuroscience and development (January 31) by Alexander Yartsev
- Rogers’ paradox: why cheap learning doesn’t raise mean fitness (February 7) by Marcel Montrey
- False memories and journalism (February 10)
- Misbeliefs, evolution, and games: a positive case (March 28) by Sergio Graziosi
- Short history of iterated prisoner’s dilemma tournaments (March 2)
- Symmetry in tag-based games & invariants under replicator dynamics (November 12)
- Diversity and persistence of group tags under replicator dynamics (November 29)
Alex shared the insights into moral decision making that he’s amassed in the last couple of years of reading. In particular, he discussed what it would mean to analyse moral behavior and explored the nascent scientific study of morality by researchers like Jon Haidt. If you enjoy this cross-road of philosophy and psychology then I would recommend the Very Bad Wizards podcast as a great supplement to Alex’s posts. Episode 3 (“We believe in nothing!” Cultural diversity, relativism, and moral truth) would be a good starting point.
Marcel discussed Rogers’ paradox in evolutionary explanations of social learning. The definition of an evolutionary equilibrium between two strategies is that the two strategies have equal fitness. So in equilibrium, social learners has to have the same fitness as individual learnings. But if the fitness of individual learns is independent of the number of social learners then in equilibrium social learning provides not net fitness benefit compared to a population of just individual learners. But then why why do populations with social learners — like humans — seem to be so much fitter than those without? Marcel presents some some solutions to Rogers’ paradox and a connection to the evolution of cooperation.
A property of our memories and learning — both social and otherwise — seems to be a systematic propensity for misbeliefs. Is this an unfortunate mistake or side-effect? Sergio suggests not, and gives some examples of adaptive misbeliefs.
Although I consider learning in my models, I seldom worry about memory. Historically, however, this has not been the case. The earliest, and most famous, applications of evolutionary game theory to the social domain involved iterated games where agents could condition their behavior on the history of their past interactions with their partner. I recall some of this history as an introduction to Vincent Knight’s iterated prisoner’s dilemma tournament.
Finally, I looked at general tag-based models to better understand behavior. Here the agents are allowed to condition their behavior not on the past history of interactions, but on an arbitrary heritable tag. Symmetries in the payoff-structure of such models when combined with my factoring trick allows us to better understand the replicator dynamics of ethnocentrism.
These 8 posts had a corpus of around 15.8 thousand words and garnered around 6.6 thousand views.