In Cooperation with






Funding
Heidelberg University, Excellence Initiative, Institutional Strategy, Instrument Scientific Computing (ZuK 5.4)

Important dates

Registration

Please register and submit a brief summary of your talk here until July 22, 2015.


Workshop Scientific Computing
in the Behavioral Sciences (SCBS 2015)




Invited speakers

  • Joscha Bach (MIT, Media Lab, USA)
    Integrating Motivational Models into Artificial Intelligence Architectures
    Human cognition is not only characterized by our ability to pursue goals, but to identify and commit to these goals. Humans are goal-finding systems. Constructing and choosing goals in an open environment requires a structured motivational system. We will discuss a possible functional abstraction of general motivation, to account for social, cognitive and physiologically motivated behavior. We will look at the relationships between motivation, cognitive modulation and affect in the resulting model. Last but not least, we will look at parametrization of the system, to account for personal variability in behavior, and opening the way to model temperament and personality traits.
  • Dietrich Dörner (University of Bamberg Department of Psychology)
    Was? Wann? Wie? Ein Modell der Absichtsregulation
    Zum gleichen Zeitpunkt haben wir meist recht viele verschiedene Motive und Absichten. Ich diktiere jetzt gerade; eigentlich müßte ich das Frühstücksgeschirr noch abdecken, eine Dissertation zu Ende lesen, usw. Die verschiedenen Absichten betreffen verschiedene Grundmotive: Durst, Ordnung im Haushalt, Affiliation, Bestimmtheit und Kompetenz; oft dient eine Absicht mehreren Grundmotiven. Es kommen auch neue Motive hinzu; Gelegenheiten und Gefahren tauchen auf: a) Es klingelt an der Tür. Ach ja, das ist Wolf. Achtung, das ist gefährlich! Mit Wolf kann man sich sehr gut unterhalten und dabei könnte ohne weiteres der Nachmittag vorbeigehen. b) Draußen signalisiert eine düstere Wolke die Gefahr eines starken Regengusses. Die Wäsche muß von der Wäschespinne. Das Absichtsprofil ändert sich also ständig. Die Stärke der Absichten nimmt zu oder ab. Neue Absichten tauchen auf, andere werden erledigt, wieder andere bleiben erstmal liegen. Ich möchte in diesem Vortrag ein mathematisches Modell zur Absichtsregulation vorstellen. Auf welche Weise kommen Absichten "an die Macht" und wie wird – je nach Absichtsprofil - umgeschaltet; vom Modus "schnell und grob“ auf den Modus "langsam und genau" oder von "vorsichtig" auf "mutig", usw. Wir validieren das Modell dadurch, dass wir eine längere Verhaltensepisode Hitlers (nämlich Hitlers Verhalten nach dem Beginn der Invasion am 6.6.1945) mithilfe des Modells erklären.
  • Björn Meder (Max Planck Institute for Human Development, MPIB, Berlin; Adaptive Behavior and Cognition)
    Optimal Experimental Design, Heuristics, and Human Information Search
    Much recent work in psychology has considered which of several optimal experimental design (OED) models, like information gain, impact, or probability gain, are best suited to the selection of information. Key issues addressed in this talk include (i) which OED model best accounts for human search behavior, (ii) whether simple heuristic strategies can approximate or even exactly implement OED models, and (iii) whether OED models, which are typically implemented in a stepwise manner (i.e., they only consider the next time step when evaluating the usefulness of different queries) are in fact optimal on sequential search tasks.
  • Magda Osman (Queen Mary, University of London)
    Past, Current and Possible Future Types of Formal Descriptions of Dynamic Decision-making Processes
    The aim here is to first try to set out the broad common research themes that have been the focus of empirical work in the domain of dynamic decision-making (DDM). From this the aim is to establish if there are connections between families of DDM models that formally capture the processes that underpin learning and decision-making in dynamic environments. Of the range of models that have been developed over the history of research on DDM (e.g., Finite State Automata, Instance-based models, ACT-R, Connectionist models), the most common assumptions made are that DDM is supported by two types of learning and memory systems, DDM improves through accumulation of action-outcome states, and through experience selective information search is achieved. One important factor that appears to be absent from most of empirical and formal work in DDM is causality, which this discussion will argue is perhaps where future DDM work should focus.
WWW-Seiten: MvS, AR