A simulation of meta-induction based prediction strategies. Meta-inductive prediction strategies are an approach that has been proposed by Gerhard Schurz to answer the problem of induction (also known as "Hume's problem" by philosophers).

Version: 0.5

(c) 2007 by Eckhart Arnold & Gerhard Schurz, MIT Open Source License

Authors: Gehard Schurz (design of the simulation and meta-inductive strategies) Eckhart Arnold (implementation in python and user interface) web: https://www.phil-fak.uni-duesseldorf.de/philo/personal/thphil/schurz/ www.eckhartarnold.de email: schurz@phil-fak.uni-duesseldorf.de eckhart_arnold@yahoo.de

In philosophy the problem of induction is understood as the problem to justify the inference of future regularities or general laws (e.g. natural laws)from observed past regularities (e.g. experimental evidence). The 18th century sceptic David Hume famously denied that induction could be justified. Most subsequent philosophers found this hard to accept and thinkers as important as Immanuel Kant and Karl Popper have proposed solutions to the problem of induction. Today, however, experts agree that none of these attempts has been successful. Therefore, the problem of induction is still an area of active research. The meta-inductivist approach, proposed by Gerhard Schurz, tries to answer the problem of induction not on the object level (e.g. induction over recurring events in the world), but on the meta level (e.g. rules for picking the best of various different inductive strategies).

In order to develop his meta-induction approach, Gehard Schurz designed computer simulations to study the characteristics of meta-inductive strategies and to design new and types of meta-inductive strategies. These simulations have been programmed by me Python code in the years between 2003 and 2007. In order to actually answer the problem of induction, it is of course necessary to mathematically prove so-called "possibility theorems" about the predictive power of meta-inductive strategies. This cannot be done by a computer simulation alone. But computer simulations are great heuristic tool in this context.

The meta-inductivist approach is discussed in-depth in the papers listed below. For a short synopsis of simulation design, see sections 2 and 3 of the extended preprint of my paper, on: http://eckhartarnold.de/papers/2009_Induktionsproblem/node2.html (I had to take these passages out for publication, because the referees considered a synposis of Schurz own presentation as unnecessary. While I agree with that, the passages might still be helpful for the beginner.)

• Schurz, Gerhard. 2003. Der Metainduktivist. Ein spieltheoretischer Zugang zum Induktionsproblem. In Proceedings der GAP.5, Bielefeld 22.-26.9.2003, ed. Roland Bluhm and Christian Nimtz. Gesellschaft für analytische Philosophie Paderborn: mentis Verlag pp. 243-257. URL: www.gap5.de/proceedings/pdf/243-257_schurz.pdf

• Schurz, Gerhard. 2008. “The Meta-inductivist's Winning Strategy in the Prediction Game: A New Approach to Hume's Problem.” Philosophy of Science 75:278-305.

• Eckhart Arnold: Can the Best-Alternative-Justification solve Hume's Problem? (On the Limits of a Promising Approach), in: Philosophy of Science 2010, 584-593, DOI: 10.1086/656010, URL (extended preprint): http://eckhartarnold.de/papers/2009_Induktionsproblem/Induktionsproblem.html

For an introduction to the problem of induction itself, I highly recommend the respective chapter in Bertrand Russell's "The problems of philosophy". Also, Nelson Goodman's "New Riddle of Induction" is very important in this context. A more recent treatment is found in Colin Howson's "Hume's problem".

Before you can run the simulation you need to install the following software on your computer:

• python2 which can by found on www.python.org . Beware, though, you need to use the old version, i.e. python2. I did not have time to port the programm to python 3, althouh I believe that this can easily be done.

• wxPython for the user interface, which can be downloaded from www.wxpython.org

• there is also a jython user interface, but it only contains very few examples. Jython is the java virtual machine version of python, see www.jython.org.

Finally, you need to clone the git repository of MetaInductionSim.

MetaInductionSim can be run, by starting either the wxPython or jython/SWING user interface. In order to start the wxPython user interface (recoommended) use the following command on the command line:

python2 Induction_Examples.py

You can pick different meta-induction scenarios from the menu then.

For the jython/SWING user interface, run:

jython InductionApplet.py

If you would like to analyze the simulation and the meta-inductive strategies by reading the source code, I suggest leaving the user interface code aside (it's pretty messy, anyway) but concentrating on the simulation code in the file

Induction.py

It is reasonably well documented with python docstrings to be understandable, I hope.

Back then when I wrote the simulation, I had to programm my own library for graph plotting: PyPlotter . Because during the last ten years python has become increasinly popular in the scientific community, there are much better plotting libraries out there, nowdays. Most probably a combination of jupyter-Notebooks jupyter.org and the matplot-Library matplotlib.org would be the best choice for anyone who would like to continue experimenting with MetaInductionSim.