1. Skip to Menu
  2. Skip to Content
  3. Skip to Footer

Talk "Newcomb's Paradox – Wissen ordnen und erschließen in hybriden Netzen" @ Foundations of Statistics (16 May, 2012)

newcomb_stats

On 16 May, 2012, Roland Poellinger gives the talk "Newcomb's Paradox – Wissen ordnen und erschließen in hybriden Netzen" at the LMU Research Seminar Foundations of Statistics (Statistics Department, Ludwigstraße 33, room 245 – 6:30pm). In various approaches to solutions of the paradox the principle of dominance and the principle of maximum expected utilities are balanced or tweaked in more or (more often) less natural ways. Reconsidering the concept of prediction as an epistemic change of state yields a compact and intuitive rendition of the problem. Excerpt from the abstract: "In diesem Vortrag möchte ich die Modellierung des Paradoxons in Bayes‘schen kausalen Modellen erläutern, wie sie von Pearl (1995 oder 2000/2009) definiert und von Wolfgang Spohn („Reversing 30 Years of Discussion: Why Causal Decision Theorists Should One-Box“) bzw. Meek & Glymour (1994) zur Analyse von Newcomb’s Problem herangezogen werden. Als Antwort auf diese Ansätze möchte ich im zweiten Teil meiner Diskussion meinen Lösungsvorschlag in Causal Knowledge Patterns (einer Erweiterung des Bayesnetz-Frameworks mit intensionalen Informationsbrücken) präsentieren, um schließlich – näher an der Intuition und der ursprünglichen Formulierung von Nozicks Geschichte – bei der Lösung des „one-boxing“ anzugelangen."

Talk "What is Causal Knowledge?" @ LMU Centrum für Informations- und Sprachverarbeitung (19 Jan, 2012)

causal_knowledge

On 19 Jan, 2012, Roland Poellinger gave the talk "What is Causal Knowledge?" at the LMU Centrum für Informations- und Sprachverarbeitung for an audience of computational linguists. The talk focused on philosophical implications of the title question, algorithmic approaches towards causal analysis, formalizations of an interventionist concept of causation, mathematical foundations of DAG search methods, and the necessity to extend the Bayes net causal model framework for the treatment of hybrid nets of mixed type knowledge (unifying extensional and intensional information).

Doctoral Thesis "Concrete Causation. About the Structures of Causal Knowledge."

diss_cover
Roland Poellinger's doctoral thesis "Concrete Causation. About the Structures of Causal Knowledge" (supervisor: Godehard Link - logic; further comittee members: Thomas Augustin, statistics, Klaus Schulz, computational linguistics, and C. Ulises Moulines, philosophy of science) centers about theories of causation, their interpretation and embedding in metaphysical-ontological questions, as well as the application of such theories in the context of science or decision theory. The main proposal is the development of a framework for integrating causal and non-causal knowledge in unified structures and the definition of methods for consistent inference of causal claims from such patterns. Applications are problems of (causal) decision theory or the representation of logical-mathematical, synonymical, as well as reductive relationships in efficiently structured, operational nets of belief propagation.

The dissertation concludes the PhD project comprising the international conference "Concrete Causation" (2010) and the appertaining channel in iTunes U containing recordings of the "Concrete Causation" conference talks and of further research.

Download a summary of the thesis here [PDF].

[Keywords: causal modeling, interventionist account of causation, DAG, Bayesian networks, intervention, causal knowledge pattern, epistemic contour, subjective causation, Newcomb's paradox, prisoners' dilemma, causal Markov condition, Judea Pearl, David Lewis, Wolfgang Spohn]

Talk "Disentangling Nets for Causal Inference" @ DGPhil 2011 (13 Sept 2011)

disentangling-talk-dgphil11
On occasion of the MCMP workshop on mathematical philosophy (13 & 14 Sep 2011) as part of the DGPhil conference 2011, Roland Poellinger gave a short presentation titled "Disentangling Nets for Causal Inference". The topics: What problems arise, when Bayes net causal models are augmented by the insertion of non-causal, non-directional knowledge? How can structural knowledge about given situations be enriched to decide about the utility of available semi-DAGs? And is there any possibility to regain the Markov condition for consistent computation of causal claims from graphs that unify directional and undirectional information? Watch the talk on iTunes U!

Talk "Reclaiming Markov in Entangled Structures of Deterministic Causal Knowledge" @ CMU Pittsburgh (30 Nov and 7 Dec 2011, 5pm)

cpk-talk-cmu

On 30 Nov 2011 Roland Poellinger will give a talk at the Center for Formal Epistemology of Carnegie Mellon University, Pittsburgh, titled "Reclaiming Markov in Entangled Structures of Deterministic Causal Knowledge – Part 1: Epistemic Contours" (5pm, Baker Hall 150). This talk will concentrate on the observation that in many cases of causal reasoning non-causal, non-directional knowledge is drawn on and computed efficiently and consistently, although reasoning with this kind of knowledge seems to violate the causal Markov condition in standard Bayes net causal models. Examples are found in causal decision theory, where modeling Newcomb’s paradox (in its original formulation) or the prisoners’ dilemma seemingly yields counter-intuitive solutions, or in cases of inter-level (e. g., genuine bottom-up or top-down) causation that are usually simply collapsed to “flat” models incorporating (unanalyzed) cross-level mechanisms. Intermediate stages in the process of reducing theories, learning, or modeling given situations (e. g., stages with intensionally separated but extensionally equal variables) find no formalized expression in causal models obeying Markov. Embedding “entangled” variables in causal models renders those models non-Markovian. If we want to stick with these models, how can Markov be reclaimed?

Part 2 of the talk, "The Principle of Explanatory Dominance", will take place on 7 Dec 2011 (5pm, Baker Hall 150).

Download the abstract here as PDF.

Talk "Newcomb's Paradox and Its Intuitively Adequate Solution in Causal Knowledge Patterns" @ Uni Melbourne (7 Oct 2011)

newcomb_in_ckps_teaser

On 7 Oct 2011 Roland Poellinger will give a talk at the University of Melbourne, titled "Newcomb's Paradox and Its Intuitively Adequate Solution in Causal Knowledge Patterns". Abstract: Referring back to the physicist William NEWCOMB, Robert NOZICK (1969) elaborates on – as he calls it – Newcomb’s problem, a decision-theoretic dilemma in which two principles of rational choice seemingly conflict each other, at least in numerous renditions in the vast literature on this topic: Dominance and the principle of maximum expected utility recommend different strategies in the plot of the game situation. While evidential decision theory (EDT) seems to be split over which principle to apply and how to interpret the principles in the first place, causal decision theory (CDT) seems to go for the solution recommended by dominance (“two-boxing”).
In this talk I will firstly present the CDT proposal by Wolfgang SPOHN (cf. “Reversing 30 Years of Discussion: Why Causal Decision Theorists Should One-Box”) who opts for “one-boxing” by employing reflexive decision graphs. The second part of my discussion – as a reply to SPOHN – will draw on the framework of causal knowledge patterns, i.e., Bayes net causal models (cf. PEARL 1995 and 2000/2009) augmented by non-causal knowledge (epistemic contours), to finally arrive at “one-boxing” – more intuitively and more closely to what actually is in NOZICK’s story.