ABSTRACT. This paper applies a cognitive modelling approach to model decision making of naïve subjects in virtual emergency situations. Virtual environments (VE) can be used as a virtual laboratory to investigate human behaviour in simulated emergency conditions. Cognitive modelling methodology and human performance data from VEs can be used to identify the problem solving strategies and decision making processes of general personnel in offshore emergency egress situations. This paper demonstrates the utility of decision trees as a cognitive tool for two main purposes: 1) assessing VE training curriculum and 2) predicting human behaviour. To show these capabilities, the results of two empirical studies are compared using a decision tree induction approach. The first experiment investigated the learning and inference process of participants trained using a lecture based teaching (LBT) approach. The second experiment used another pedagogical approach – simulation-based mastery learning (SBML). Overall, decision trees were found to be a useful method for evaluating the efficacy of VE training, and as a basis for predicting individuals’ decision-making performance.

ABSTRACT. In many transportation modes, automation is added to increase comfort, efficiency, or to reduce human errors. Automation has a direct impact on the drivers workload, which can even be higher then without automation. In this paper we propose the development of a virtual driver that can predict human workload in early design phases of automation and assistant systems. We describe the workload model in a closed-loop simulation and an early validation.

ABSTRACT. Functional cognitive models are used to explain observed human behavior. Applying such models to predict behavior requires generalization of the model to be applied in different application domains but also a careful consideration of model input data validity. Visual attention models have already been validated in various domains. But elicitation techniques to collect valid input data that is reproducible by others are still missing. For visual attention prediction model input data is determined mainly based on discussion between experts and individual experience, which is difficult to reproduce. We use a software tool to ensure input validity. The tool helps users to create attention models. It uses images of the situations that are investigated for stimulating the imagination of users to put themselves into these situations. An experiment (n=40) showed that using looping videos instead of static images stimulates imagination in a different way. It has an effect on the models generated by the users and needs careful consideration.

ABSTRACT. A modeling approach addressing visual search in an array of items of differing similarity is introduced. The model is able to capture the effects found in a study that varies target-distractor similarity (low vs. high), distractor-distractor similarity (low vs. high) of icons, target presence (present vs. absent) and the set size (8, 16 or 24 icons). To be able to simulate human visual search in such a task with original ACT-R mechanisms we implemented a hybrid search strategy that combines parallel and serial search. The presented model can provide useful insight for researchers interested in modeling tasks containing visual icon search.

ABSTRACT. This paper describes an experiment and fuzzy set models in the domain of linguistic labels for simple spatial relationships: for example, that one object is "in front of" or "to the right of" another. Input to the models was generated by robot sensors (camera and distance sensors), from a viewer perspective on different configurations of two objects. Performance of the models is is qualitatively similar to human judgments; performance is also quantitatively similar to that of a model working from an environmental bird's-eye view. Such models are one component of a robot’s interpretation of the context of human activity.

ABSTRACT. Generating truly random sequences is hard. When participants are engaged in a competitive game (e.g., Matching Pennies), the sequences they generate are surprisingly more random than when given explicit instructions to generate random sequences (Rapoport and Budescu, 1992). To explore this phenomenon, we formalized two probabilistic models of Theory of Mind reasoning about subjective randomness. One model (the Fair-Coin model) assumes participants predict their opponents’ choices by implicitly assuming that their opponents intend to generate binary sequences that simulate the outcome of tossing a fair coin. The other model (the Markov model) assumes participants believe that their opponents intend to generate sequences that simulate the outcome of a Markov process with transition probability equal to 0.5. We find that Theory of Mind models of both the Fair-Coin and the Markov definitions of subjective randomness are able to characterize the calibrated subjective randomness that occurs when participants are playing an iterated competitive game (Matching Pennies), but the Markov Model is better than the Fair-Coin Model in simulating the situation where participants need to specify their choice sequences in advance of the game. The current study suggests that the calibrated subjective randomness in competitive games can be explained by the online evaluation of sequence randomness with Theory of Mind reasoning.

ABSTRACT. Detailed transfer of procedural knowledge has been modeled in Actransfer, an extension of ACT-R, by combining the primitive memory operations of productions (PRIMs) with the architecture's procedural learning mechanism \citep{taatgen2013nature}. This work explores whether these same principles can be applied to the Soar cognitive architecture, which uses different models of working memory and procedural learning. We confirm that these principles can transfer to an unmodified version of Soar. Our research contributes a novel model of skill learning based upon a deeper level of primitive skill composition than described in the PRIM model that is suitable for unbounded working memory architectures, and which yields transfer profiles similar to those revealed in human studies.

ABSTRACT. Term rewriting is a well established formal method used for defining semantics of programming languages, program transformations, automatic theorem proving, symbolic programming, intelligent tutoring system development etc. In this paper, we present a language based on term rewriting as an alternative formalism for modelling cognitive skills. We show how the language overcomes some deficiencies of production systems (compositionality, readability, control-flow etc.) and how, as a consequence, it can help with addressing practical problems raised by the cognitive modelling community.

ABSTRACT. Speakers often refer to context only implicitly when using language. "It's warm outside" could signal warm relative to other days of the year or just the current season (e.g., warm for winter). "Warm" conveys the temperature is high relative to some comparison class, but little is known about how a listener decides upon such a standard of comparison. We formalize how world knowledge and listeners' internal models of speech production can drive the resolution of a comparison class in context. We introduce a Rational Speech Act model and derive two novel predictions from it, which we validate in an experiment that measures listeners' beliefs about the likely comparison class used by a speaker. Our model makes quantitative predictions given prior knowledge for the domains in question. We triangulate this knowledge with a follow-up language task in the same domains, using Bayesian data analysis to infer priors from both data sets.

ABSTRACT. How do humans learn the syntax and semantics of words from language experience? How does the mind discover abstract relationships between concepts? Computational models of distributional semantics can analyze a corpus to derive representations of word meanings in terms of each word’s relationship to all other words in the corpus. While these models are sensitive to topic (e.g., tiger and stripes) and synonymy (e.g., soar and fly), the models have limited sensitivity to part of speech (e.g., book and shirt are both nouns). By augmenting a holographic model of semantic memory with additional levels of representations, we present evidence that sensitivity to syntax is supported by exploiting associations between words at varying degrees of separation. Our hierarchical holographic memory model bridges the gap between models of distributional semantics and unsupervised part-of-speech induction algorithms, providing evidence that semantics and syntax exist on a continuum and emerge from a unitary cognitive system.

ABSTRACT. This paper investigates how cognition facilitates the adoption of new words through a study of the large-scale Reddit corpus, which contains written, threaded conversations conducted over the internet. Parameters for the cognitive architecture are estimated. Using ACT-R's account of declarative memory, the activation of memory chunks representing words is traced and compared to usage statistics sampled from a year of data. Potential values for decay and retrieval threshold are identified according to model fit and growth rates of word adoption. The resulting estimate for the decay parameter, d, is 0.22, and the estimate for the retrieval threshold parameter, rt, lies between 3.4 and 4.5.

ABSTRACT. We examine working memory use and incrementality using a cognitive model of grammatical encoding. Our model combines an empirically validated framework, ACT-R, with a linguistic theory, Combinatory Categorial Grammar, to target that phase of language production. By building the model with the Switchboard corpus, it can attempt to realize a larger set of sentences. With this methodology, different strategies may be compared according to the similarity of the model's sentences to the test sentences. In this way, the model can still be evaluated by its fit to human data, without overfitting to individual experiments. The results show that while having more working memory available improves performance, using less working memory during realization is correlated with a closer fit, even after controlling for sentence complexity. Further, sentences realized with a more incremental strategy are also more similar to the corpus sentences as measured by edit distance. As high incrementality is correlated with low working memory usage, this study offers a possible mechanism by which incrementality can be explained.