Professorship Predictive Analytics
Nexus Cognitive Modeling

Cognitive Modeling

 

Welcome

Welcome to our collaboration network! We are a worldwide, informal network of scientists working on cognitive modeling. In regular (online) events, we share our research and discuss different modeling approaches. Feel free to join our network!

Keep up to date with our mailing list

You are welcome to subscribe our mailing list 'cognitive-modeling'. We will share upcoming events and other relevant information. To subscribe the mailing list, send a mail with "subscribe" to .

 

Organization

We regularly organize educational/tutorial and research talks focused on cognitive modeling. Are you interested in presenting and discussing your research topic or modeling approach in our collaboration network? You are welcome to !

Previous Talks

Date Topic Referent
15.06.2022
Social motorics - a predictive processing model for efficient embodied communication

Social motorics - a predictive processing model for efficient embodied communication

Social interaction is prone to misunderstandings, yet, its reciprocal processes enable robust recovery and allow for successful belief coordination, i.e., making sure you and your interaction partner know what you are talking about.

We present a hierarchy of nested generative sensorimotor perception and action processes based on the free-energy principle and active inference, which sees action as a form of inference over possible ways to make the environment meet the agent's predictions. These principles are used to simulate the interplay of core processes within the so-called social brain, i.e., identified sensorimotor processing and mentalizing areas in the human brain, associated with activity during social interaction. Building on recent work on computational models of agency, sensorimotor- and mentalizing processes, we propose a computational hierarchy that contributes a bootstrapped account of mentalizing. This enables a process for belief coordination similar to the we-mode proposed by Frith (2012): An implicit form of mentalizing that bootstraps the attribution of mental states that seems automatic, where contextual information and prior information can influence behavioral understanding top-down.

We argue that nested high-level representations that enable belief coordination do not only act to minimize uncertainty during perception and action, but they also assume a form of metacognitive regulatory control based on the uncertainty detected in the social interaction at large. In other words, prediction errors during inference of social information may lead to repairs and other reciprocal communication behavior, which may be integral to communication in general.
First, we introduce the foundational modeling approach and then show simulation results between three agents in a leader-follower non-verbal communication game. The agents perform reciprocal belief coordination that involves regulation of the whole nested hierarchy, its sensorimotor and mentalizing parts. The results stress the importance of balancing prior beliefs with new information by means of a precision-weighting bias that regulates not only perception and action, but also more generally, the gain on social information.

Dr. Sebastian Kahl
08.06.2022
Reasoning for Commonsense Knowledge

Reasoning for Commonsense Knowledge

Associative reasoning refers to the human ability to focus on knowledge that is relevant to a particular problem. In this process, the meaning of symbol names plays an important role: when humans focus on relevant knowledge about the symbol 'ice', similar symbols like 'snow' also come into focus. In this talk, we discuss different ways to model this associative reasoning by introducing selection strategies that extract relevant parts from large commonsense knowledge sources. These selection strategies are based on word similarities from word embeddings and are therefore able to take the meaning of symbol names into account. We demonstrate the usefulness of these selection strategies with results from theorem proving on commonsense problems and with a case study from creativity testing.

 Prof. Dr. Ulrich Furbach, Dr. Claudia Schon
01.06.2022 Evidenzakkumulationsmodelle und die Modellierung von kognitiver Beanspruchung bei geteilter Aufmerksamkeit Daniel Trommler
25.05.2022
The Relevance of Formal Logics for Cognitive Logics, and Vice Versa

The Relevance of Formal Logics for Cognitive Logics, and Vice Versa

Classical logics like propositional or predicate logic have been considered as the gold standard for rational human reasoning, and hence as a solid, desirable norm on which all human knowledge and decision making should be based, ideally. For instance, Boolean logic was set up as kind of an arithmetic framework that should help make rational reasoning computable in an objective way, similar to the arithmetics of numbers. Computer scientists adopted this view to (literally) implement objective knowledge and rational deduction, in particular for AI applications. Psychologists have used classical logics as norms to assess the rationality of human commonsense reasoning. However, both disciplines could not ignore the severe limitations of classical logics, e.g., computational complexity and undecidedness, failures of logic-based AI systems in practice, and lots of psychological paradoxes. Many of these problems are caused by the inability of classical logics to deal with uncertainty in an adequate way. Both disciplines have used probabilities as a way out of this dilemma, hoping that numbers and the Kolmogoroff axioms can do the job (somehow). However, psychologists have been observing also lots of paradoxes here (maybe even more).

So, are humans hopelessly irrational? Is human reasoning incompatible with formal, axiomatic logics? In the end, should computer-based knowledge and information processing be considered as superior in terms of objectivity and rationality?

Cognitive logics aim at overcoming the limitations of classical logics and resolving the observed paradoxes by proposing logic-based approaches that can model human reasoning consistently and coherently in benchmark examples. The basic idea is to reverse the normative way of assessing human reasoning in terms of logics resp. probabilities, and to use typical human reasoning patterns as norms for assessing the cognitive quality of logics. Cognitive logics explore the broad field of logic-based approaches between the extreme points marked by classical logics and probability theory with the goal to find more suitable logics for AI applications, on the one hand, and to gain more insights into the rational structures of human reasoning, on the other.

Prof. Dr. Gabriele Kern-Isberner
18.05.2022
Balancing control: a Bayesian interpretation of habitual and goal-directed behavior.

Balancing control: a Bayesian interpretation of habitual and goal-directed behavior.

Arbitrating between fast automatic behavior and slow but adaptive goal-directed behavior allows humans to allocate resources in a situation-appropriate manner. However, how this is achieved is still an open research question. In this talk I will present a Bayesian cognitive decision making model which implements hypothesized mechanisms for this arbitration: Goal-directed control is based on an explicit representation of action-outcome contingencies, automatic behavior is encoded in a prior which implements an a priori bias to repeat actions or sequences, and both are learned in a context-dependent manner. Control is natively arbitrated using Bayes' theorem. I will show using simulations that this model replicates key features of habit learning, as a case of automatic behavior, as well as typical results in two cognitive control tasks. These results suggest that the underlying mechanisms may be more general control features that underlie a wide range of human behaviors.

 Dr. Sarah Schwöbel
11.05.2022
Risk and consistent decision making

Risk and consistent decision making

Decision-making under uncertainty involves uncertainties – and this talk addresses them from a mathematical perspective: how to quantify uncertainties and how to handle and manage unforeseen events? Evolving systems require even consecutive decision-making. In these situations, it is essential to design decisions to reach the goal specified and to avoid or reduce corrective actions. We elaborate the mathematical framework and address measures, which allow assessing the evolution of the process over time.

 Prof. Dr. Alois Pichler
04.05.2022
Modeling Human Reasoning: Benchmarking, Analysis, and Improvement

Modeling Human Reasoning: Benchmarking, Analysis, and Improvement

The field of human reasoning research can look back on over a century of interdisciplinary work aimed at uncovering the cognitive processes underlying the human ability to make inferences. However, despite this extensive history, recent investigations suggest flaws in the methodological approach predominantly employed by the field. For instance, the traditional focus has been on aggregate analyses, i.e., trying to understand and explain group-level behavior that neglects the importance of individual differences. This invites the problem of group-to-individual generalizability, which may render the transfer of insight to individual behavior invalid. If proven to hold for the field of reasoning research, current theories would not be applicable to individual human reasoners but only to artificial, average behavior with limited scientific relevance.

The present thesis presents a novel approach to reasoning research that aims to provide a remedy for the current shortcomings of the field. This is accomplished by shifting the methodological perspective from the prevailing focus on groups to individuals. In particular, the fundamental research goals are reformulated as a series of predictive modeling problems that naturally invite the use of accuracy metrics to gauge the performance of models. By jointly allowing theory-driven and data-driven approaches to compete on a common ground, this new perspective does not only allow for an improved assessment of performances, but also for detailed analyses of the shortcomings and advantages of specific approaches, which may boost synergistic effects in the field.

Using this novel approach, syllogistic and spatial relational reasoning, two central domains in the field of reasoning research, are analyzed. Benchmarks consisting of collections of state-of-the-art models that reflect the theoretical progress in the field, as well as a selection of representative datasets, are created to run the field's most comprehensive model evaluations to date. The results suggest that the problems of group-to-individual generalizability are present in the field. As such, the prevailing neglect of inter-individual differences that has caused a focus on "average" instead of individual behavior leads to the risk of theories losing their explanatory relevance. Furthermore, practical shortcomings of the data landscape such as noise and the lack of discriminability between individuals are uncovered. The thesis concludes by proposing an improved methodological paradigm for approaching model-driven research in the field.

 Nicolas Riesterer
20.04.2022
Humans Reason Skeptically

Humans Reason Skeptically

The weak completion semantics is a novel cognitive theory. It is multi-valued, non-monotonic, knowledge rich, allows learning, can handle inconsistent background knowledge, and can be applied to model the average reasoner. Moreover, it uses abduction to explain observations, to satisfy integrity constraints, and to search for counterexamples. In all these applications, human reasoning tasks can only be adequately modelled within the weak completion semantics if skeptical abduction is applied, rather than credulous abduction. This will be illustrated in the context of the suppression task, disjunctive reasoning, and conditional reasoning.

 Prof. Dr. Steffen Hölldobler
19.04.2022
SlimStampen: Using Cognitive Computational Memory Models to Improve How 1.2 Million Dutch Students Learn

SlimStampen: Using Cognitive Computational Memory Models to Improve How 1.2 Million Dutch Students Learn

In this talk, I will discuss the SlimStampen adaptive learning system. SlimStampen, based on state-of-the-art computational cognitive models of the human memory system, tracks the memory performance of students while learning factual materials, and based on the derived internal model optimises the scheduling of to be learned items. Using this system, learning efficiency increases with 10 to 20%. In collaboration with large publishing houses, SlimStampen is now widely used in The Netherlands, resulting in rich datasets which, coming full circle, inform theories on the human memory system. In this talk, I will discuss the outline of the system, how it is used, and our most recent findings including the link with neuroscience, how speech-parameters can be used to optimise learning, and how gigabytes of learning data will change our view on how information is forgotten at short and longer durations.


Please, use this conference room.		 Prof. Hedderik van Rijn
02.02.2022
Towards inter-individually generalizing machine learning models of brain function

Towards inter-individually generalizing machine learning models of brain function

One rarely explicitly stated central assumption of most group level analyses is that human brains have similar functional anatomical organization. Contrary to this assumption, machine learning models trained on data from one brain typically do not generalize well to data measure with other individual's brains. This talk is about our recent approaches to generate machine learning models of human brain function that relax this the strong assumption of a tight coupling between brain anatomy and function. I will demonstrate that this improves generalization of the models between individual brains.

 Prof. Dr. Jochem Rieger
19.01.2022
Event-Predictive Cognitive Modeling

Event-Predictive Cognitive Modeling

Over the last decades much research evidence suggests that our brain develops event-predictive models of its environment and uses these models to interact with the world in a self-motivated, goal-directed manner. In this talk I will present recent behavioral evidence, that implies how our mind uses its event-predictive models to guide behavior. Moreover, I will introduce parts of out cognitive modeling work, which mimic this behavior qualitatively and quantitatively. I will start with models of infant behavior, proceed with models of the anticipatory cross-modal congruency effect, and end with models of event-predictive language use and social inference.

 Prof. Dr. Martin V. Butz
12.01.2022
On Uncertainty and Inconsistency in Knowledge Representation

On Uncertainty and Inconsistency in Knowledge Representation

In this talk, I will discuss recent works addressing challenges on handling uncertainty and inconsistency in knowledge representation. In particular, I will talk about formal argumentation as an approach to non-monotonic reasoning that explictly considers the roles of arguments and counterarguments in reasoning. In this context, I will briefly present recent works dealing with probabilistic uncertainty and algorithmic issues. Moreover, I will talk about inconsistency measurement as an analytical tool to assess the severity of inconsistencies in knowledge representation formalisms.

 Prof. Dr. Matthias Thimm
05.01.2022
Context sensitive anticipation of the cognitive state in human machine interactions

Context sensitive anticipation of the cognitive state in human machine interactions

For a successful collaboration between humans and intelligent systems mutual understanding of the task and environmental conditions is required. In addition, understanding the goal and expectations of the partner also needs to be considered. To address all these different aspects, joint interdisciplinary approaches are necessary to achieve this goal.
For anticipating the partner, humans use internal models and shared representations of the situation and tasks. Cognitive architecture provide means to anticipate perceptual and cognitive processes, to predict the underlying expectations and decision in dynamically evolving situations.
A few examples will be shown that provide first approaches of dynamic anticipation of cognitive processes and situation understanding that can be aligned with further information about the human partner such as motor actions, gazes, and ERPs (event related potentials).
These examples address complex interaction situations such as pilots experiencing different kind of warnings in the cockpit, drivers of highly autonomous cars that need to take over control and users learning how to interact with new mobile applications.

 Prof. Dr.-Ing. Nele Rußwinkel
30.11.2021
Memory Control in Working and Long-Term Memory

Memory Control in Working and Long-Term Memory

Forgetting has a bad reputation. Yet, intentional forgetting is essential to deal with the vast amount of information we are confronted with each day – it helps us to distinguish between relevant and irrelevant information. Various studies have established that humans can intentionally forget previously learned information. When people are told to remember or forget words, memory for to-be-remembered (TBR) words is typically better than for to-be-forgotten (TBF) ones. Despite decades of research on this topic, there is an ongoing debate on which cognitive mechanisms underly these directed forgetting (DF) effects: What happens with TBF information in memory when it is no longer needed and does the intent to remember boost cognitive representations? In addition, forgetting should be crucial when trying to overcome old habits. Yet, while many studies have demonstrated DF for declarative memory (e.g., lists of words), the role of intentional forgetting in action control (e.g., procedural memory) is less clear. In this talk, I will present work that aims to investigate both aspects. First, I will show evidence for DF in working memory and discuss its long-term effects to pinpoint some of the mechanisms underlying intentional forgetting in declarative memory. Second, I will transfer this knowledge to answer the question of how DF affects the formation and/or retrieval of associative memory, in particular stimulus-response associations. Last, moving beyond explicit memory instructions, I will discuss the role of memory control in the processing of errors and when inhibiting one’s actions. Together, this work underlines our ability as well as its limitations to flexibly up- or downregulate the strengths our different types of memory representations.

 Hannah Dames
24.11.2021
A comparison between Machine Learning and Cognitive Models on Individual Human Reasoning in Optimal Stopping Problems

A comparison between Machine Learning and Cognitive Models on Individual Human Reasoning in Optimal Stopping Problems

Finding the highest or lowest value in sequentially presented options with no way of returning to a previously seen option and not knowing what the future options will look like, is a task that we face regularly e.g. while searching online for a plane ticket for the next vacation. Such a task is called an optimal stopping problem. This master’s thesis considers the state of the art cognitive models for modelling human reasoning in optimal stopping problems and adapts them to make predictions on the individual level i.e. how good are models in predicting when a reasoner decides for a presented option. These cognitive models are compared to several machine learning based models to evaluate if the absence of an underlying assumption of the reasoning process, which the cognitive models have, gives the machine learning models an advantage. The models are evaluated on two different optimal stopping problems with the machine learning models outperforming the cognitive models by roughly three percentage points predictive accuracy in both problems and reaching a maximum predictive performance of 92.62%.

 Manuel Guth
12.11.2021
Theory of Mind and Epistemic Planning for Human-Robot Collaboration

Theory of Mind and Epistemic Planning for Human-Robot Collaboration

Epistemic planning is a branch of automated planning within AI where planning agents explicitly reason about the mental states of other agents: their beliefs, knowledge, plans and goals. Essentially the idea is to give the planning agents a Theory of Mind to be able to take the perspective of other agents. The talk will introduce logical models (models of dynamic epistemic logic, DEL) to represent a Theory of Mind in robots, and illustrate how this has been used to make robots pass false-belief tasks and be proactively helpful by reasoning about the goals and (potentially false) beliefs of human agents. The goal of equipping robots with a Theory of Mind is to allow more fluent and natural human-robot collaboration. A key issue is implicit coordination: how to successfully achieve joint goals in decentralised multi-agent systems without prior negotiation or coordination, but exclusively through the ability to take the perspective of the other agents.

 Prof. Thomas Bolander, Ph.D.
10.11.2021
Prediction of reaction times in a hand motor task based on oscillatory features of the electroencephalogram

Prediction of reaction times in a hand motor task based on oscillatory features of the electroencephalogram

Repeating the same motor task multiple times, we typically observe varying task performances even though experimental side conditions are kept as constant as possible. In the context of sports science, for rehabilitation after brain damage and for the use of brain-computer interfaces it may be interesting to know, if brain signals exist which are informative about the task performance. We conduced a study with healthy users and chronic stroke patients executing an isometric hand force task, and measured multiple performance metrics on a single-trial basis. Analysing the subjects' ongoing electroencephalogram (EEG) signals with machine learning methods, we found that oscillatory neural markers can be decoded from the multichannel EEG recordings, which not only are informative about the reaction time and other performance metrics in single trial, but which can also temporally predict the expected performance a few hundred milliseconds prior to task start. This research may add to the existing literature on near-threshold sensory stimulus perception, which is known to correlate with bandpower and the phase of neural signals. The possibility to predict motor performance may open the door to novel closed-loop training approaches in the medical and non-medical field.

 Dr. Michael Tangermann
03.11.2021
Nachwuchsforscherprojekt KogSys: Erkennung von Hilfsbedürftigkeit mittels Memory Hidden Markov Models

Nachwuchsforscherprojekt KogSys: Erkennung von Hilfsbedürftigkeit mittels Memory Hidden Markov Models

Die Nachwuchsforschergruppe „Sozial agierende, kognitive Systeme zur Feststellung von Hilfsbedürftigkeit” (kurz KogSys, 2016-2019, https://www.tu-chemnitz.de/informatik/KI/projects/social/) verfolgt die Entwicklung eines sozial-kognitiven Systems zur Erkennung von Hilflosigkeit. Durch die Kombination einer fortschreitenden Automatisierung und Fortführung der aktuellen demografischen Entwicklung ergibt sich die gesellschaftliche Herausforderung, älteren Menschen möglichst lange ein selbstbestimmtes Leben und die Teilhabe am sozialen Leben zu ermöglichen. Auf Basis neuster Erkenntnisse der Künstlichen Intelligenz sollen daher Maschinen in die Lage versetzt werden, den Menschen besser zu verstehen und sich an den Menschen anzupassen. Als ersten wesentlichen Schritt in diese Richtung soll die Hilfsbedürftigkeit eines Menschen bei der Bedienung von Systemen erkannt werden. Im Rahmen dieses Vortrags wird das Teilprojekt der Physik vorgestellt, in dem die Erkennung von Hilfsbedürftigkeit mittels Memory Hidden Markov Models realisiert wurde. Dazu wurden sensorische Information wie Mimik, Körperhaltung und Bewegung, aber auch die Bedienung des Gerätes herangezogen. Der hierfür verwendete Datensatz aus der ersten Studie des Nachwuchsforscherprojekts wird ebenso vorgestellt.

 Kim Schmidt
20.10.2021
Sensorische Detektion/Prädiktion von Diskomfort im automatisierten Fahren

Sensorische Detektion/Prädiktion von Diskomfort im automatisierten Fahren

Komfortables automatisiertes Fahren wird als eine wesentliche Bedingung für die breite Akzeptanz dieser Technologie betrachtet. Im automatisierten Fahren können neue potentiell diskomfortauslösende Faktoren entstehen wie z.B. die Vorhersagbarkeit und Natürlichkeit von Fahrmanövern. Eine möglichst frühe sensorische Erkennung von entstehendem Diskomfort würde die Basis für situationsbezogene Anpassungen des automatisierten Fahrstils schaffen, welche das automatisierte Fahrerleben verbessern und Eskalationen wie z.B. Reisekrankheit vermeiden könnten. Im Projekt KomfoPilot (http://bit.ly/komfopilot) wurden an der TU Chemnitz mehrere Studien zu dieser Thematik im Fahrsimulator und auf einem Testgelände mit insgesamt über 100 Probanden durchgeführt. Zum Einsatz kam verschiedenste Sensorik wie ein Handregler für eine kontinuierliche Rückmeldung von Diskomfort, Smartbands für physiologische Daten wie Herzrate und Hautleitwert, Eye Tracking, Motion Tracking, eine Sitzdruckmatte und videobasiertes Face Tracking. Im Vortrag wird ein Überblick über die Methodik, die verfügbare Datenbasis sowie bisherige Ergebnisse gegeben - mit einem Ausblick und Diskussion möglicher weiterführender Arbeiten mit diesem Datensatz.

 Dr. Matthias Beggiato

 

Get together with other modelers

Below we offer an overview of events, conferences and workshops that might be interesting for modelers.

Conferences:

Workshops:

 

Our collaboration network

The map shows other institutes and colleagues who are working on cognitive modeling.

 

Variety of Modeling

Here, we collect interesting publications for different modeling approaches.

Website: http://act-r.psy.cmu.edu/publication/

Website: https://orca.informatik.uni-freiburg.de/ccobra/

Modeling of linguistic terms, e.g. for application in verbal questionnaire scales and for intercultural research

  • Bocklisch, F. (2019). An Different or the Same? Determination of Discriminatory Power Threshold and Category Formation for Vague Linguistic Frequency Expressions. Frontiers in Psychology, 10, https://doi.org/10.3389/fpsyg.2019.01559
  • Bocklisch, F., Georg, A., Bocklisch, S.F. & Krems, J.F. (2013). Do you mean what you say? The effect of uncertainty avoidance on the interpretation of probability expressions - A comparative study between Spanish and German. In Knauff, M. Pauen, N.Sebanz, & I. Wachsmuth (Eds.), Proceedings of the 35th Annual Conference of the Cognitive Science Society (pp. 1917-1922). Austin, TX: Cognitive Science Society
  • Bocklisch, F., Bocklisch, S. F., & Krems, J.F. (2012). Sometimes, often, and always: Exploring the vague meanings of frequency expressions. Behaviour Research Methods, 44(1), 144-157. https://doi.org/10.3758/s13428-011-0130-8

Self-learning fuzzy algorithms for driver intention recognition

  • Bocklisch, F., Bocklisch, S.F., Beggiato, M., & Krems, J. F. (2017). Adaptive fuzzy pattern classification for the online detection of driver lane change intention. Neurocomputing. https://doi.org/10.1016/j.neucom.2017.02.089

Fuzzy classifier sequences for diagnostic reasoning (e.g., in medicine) and modeling of human expert knowledge

  • Bocklisch, F. & Hausmann, D. (2018). Multidimensional fuzzy pattern classifier sequences for medical diagnostic reasoning. Applied Soft Computing, 66, 297-310. https://doi.org/10.1016/j.asoc.2018.02.041
  • Bocklisch, F., Stephan, M., Wulfken, B., Bocklisch, S.F., & Krems, J.F. (2011). How Medical Expertise Influences the Understanding of Symptom Intensities – A Fuzzy Approach . In A. Holzinger and K.-M. Simonic (Hrsg.), Information Quality in e-Health: USAB 2011, LNCS 7058 (pp. 703-706). Springer: Heidelberg.

  • Schmidt, K., & Hoffmann, K. H. (2019). Modified Baum Welch Algorithm for Hidden Markov Models with Known Structure. In International Conference on Intelligent Human Systems Integration (pp. 497-503). Springer, Cham.

Website: https://www.nengo.ai/

Website: https://www.ai.rug.nl/~niels/prims/index.html

Portrait: Prof. Dr. Dr. Marco Ragni
Prof. Dr. Dr. Marco Ragni
  • Phone:
    +49 371 531-38284
  • Email:
  • Address:
    Reichenhainer Straße 41, 09126 Chemnitz
  • Room:
    C47.508 (prev. 2/41/508)
Portrait: M. Sc. Sara Todorovikj
M. Sc. Sara Todorovikj
  • Phone:
    +49 371 531-31653
  • Email:
  • Address:
    Reichenhainer Straße 41, 09126 Chemnitz
  • Room:
    C47.501 (prev. 2/41/501)
Portrait: M. Sc. Daniel Trommler
M. Sc. Daniel Trommler

University News