Experiments

Teaching children  to attribute second-order false beliefs: A training study

We designed a new experiment in which we trained children of ages 5 and 7 with second-order false belief stories to investigate which kind of feedback (if any) will help them. Second-order false belief stories are stories in which the children are asked to attribute a second-order belief such as “Marja thinks that Ken thinks that she took potatoes”, where the children know that Marja’s belief is in fact false. For this purpose, we constructed 34 stories with related pictures. All of the pictures have been designed by the professional artist Avik Kumar Maitra. This link shows one of the stories and the related pictures.

Modeling developmental transitions in reasoning about false beliefs of others

We constructed an ACT-R model in order to show the developmental transitions in reasoning about false beliefs of others. These stand for a child’s reasoning from his/her own point of view (zero-order) to taking into consideration an other agent’s beliefs (first-order) and later to taking into consideration an other agent’s beliefs about other agents’ beliefs (second-order). You can find the model code here.

Marble Drop game to investigate adults’ second-order theory of mind

One paradigm that we currently use to investigate higher-order social cognition is the Marble Drop game. You can read about it here.

Using computer agents to discover the benefit of higher-order theory of mind reasoning

It is difficult to control how much theory of mind a human participant can use in an experimental setting. This makes it hard to experimentally determine the effectiveness of higher-order theory of mind reasoning across different settings. Instead, we can make use of computational agents that are designed to differ in their theory of mind abilities. Simulations using these agents show that the use of higher-order theory of mind can be beneficial to agents. To show how this benefit works, there are a number of applets and scripts on this site that simulate interactions between computer agents in different settings. These applets show how higher-order theory of mind can benefit agents in competitive settings such as rock-paper-scissors or Limited Bidding. But agents can also benefit from the use of higher-order theory of mind in a more cooperative settings such as the Tacit Communication Game. However, it turns out that higher-order order theory of mind plays an important role in mixed-motive settings, where cooperative and competitive aspects both play a role. An example of such a mixed-motive setting is negotiation in Colored Trails, both in single-shot negotiation settings and multi-round negotiation settings.

In negotiation settings, theory of mind can benefit agents as well. In an online appendix to the paper “Higher-order theory of mind in negotiations under incomplete information”, we show the mathematical details of how theory of mind agents function in the game of Colored Trails. This appendix can be downloaded here.

Evolution of altruistic punishment in heterogeneous populations

Evolutionary models of cooperation generally assume that groups consist of homogeneous individuals. In these models, each individual experiences the same risk associated with cooperation, has the same benefits from successful cooperation, and is equally capable of punishing those that fail to cooperate. This applet shows how heterogeneity among individuals can affect cooperative behavior in a public goods game.

MSc Projects

Where does Jack think that Mary will look for the chocolate? Investigating the reasoning steps while answering questions about other agents, in an eye-tracking study

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl

Burcu Arslan, e-mail b.arslan@rug.nl

Adults are good at reasoning about an agent who is reasoning about another agent (e.g. “John thinks that Mary thinks that there is an apple on the table”). However, we do not know the steps of the process of reasoning that adults go through (see Arslan, Taatgen, & Verbrugge, 2013 for a computational model that predicts children’s reasoning steps).

We constructed 30 different Dutch stories to test participants’ reasoning abilities about other agents’ mental states. In the planned project, adult participants will listen to these stories with the help of headphones and watch the related drawings on the computer. They will be expected to answer a few questions about each of the stories (see Arslan, Verbrugge, Taatgen, & Hollebrandse for a study with children).

In this project, you will investigate adults’ eye-movements while they are answering questions in which they are supposed to reason about other agents’ mental states. You are expected to write a MATLAB code to run the experiment, to find 25 participants between the ages 18 and 35 to conduct the experiment, to analyse the eye-tracking data, and to write a report.

References

Arslan, B., Taatgen, N. A., & Verbrugge, R. (2013). Modeling Developmental     Transitions in Reasoning about False Beliefs of Others. In R. West & T. Stewart (eds.), Proceedings of the 12th International Conference on Cognitive Modeling, Ottawa:Carleton University. 77-82. (http://www.ai.rug.nl/~barslan/resources/Arslan_Taatgen_Verbrugge_2013.pdf)

Arslan, B., Verbrugge, R., Taatgen, N. & Hollebrandse, B. (2014). Teaching children to attribute second-order false belief: A training study. In Szymanik, J. & Verbrugge, R. (eds.), Proceedings of the Second Workshop Reasoning About Other Minds: Logical and Cognitive Perspectives, co- located with Advances in Modal Logic, Groningen, The Netherlands, 1208, pp. 1-5.(http://www.ai.rug.nl/~barslan/resources/Arslan_et_al._2014.pdf)

Modeling strategies for turn-based games

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl

Stedan Wierda, e-mail Wierda.stefan@gmail.com

Theory of mind, the ability to reason about other people’s knowledge and plans, can be studied using a wide range of tasks. One of these tasks, namely playing strategic games, has been utilized quite frequently in these studies. A marble drop game (cf. Figure 1) — a two-player turn-based game developed at our AI institute in Groningen — is one such.

Recently, we modeled two possible strategies for players in this game in ACT-R and ACTransfer (an extension of ACT-R). However, you can imagine that a large variety of strategies are possible for players involved in a game like marble drop. We have formulated a list of possible strategies that could be used to play the game, taking into account some of the constraints of the human mind. The goal of this project is to implement the formalized strategies into ACT-R models. Consequently, the reaction times predicted by the models will be tested against existing experimental results from previous studies performed at our institute. This will aid in the investigations of the usage of formal methods in cognitive studies.  For this project, some experience with the ACT-R cognitive architecture is required and coding skills form a pre-requisite.

References

B. Meijering, H. van Rijn, N.A. Taatgen and R. Verbrugge, What eye movements can tell about theory of mind in a strategic game. PLoS ONE, 7 (9) (2012), e45961. http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0045961

S. Ghosh, B. Meijering and R. Verbrugge, Strategic reasoning: Building cognitive models from logical formulas. Journal of Logic, Language and Information, 23(1) (2014) 1-29. http://link.springer.com/article/10.1007/s10849-014-9196-x#page-1

Bachelor Projects

Where does Jack think that Mary will look for the chocolate? Investigating the reasoning steps while answering questions about other agents in an eye-tracking study

Adults are good at reasoning about an agent who is reasoning about another agent (e.g. “John thinks that Mary thinks that there is an apple on the table”). However, we do not know the steps of the process of reasoning that adults go through (see Arslan, Taatgen, & Verbrugge, 2013 for a computational model that predicts children’s reasoning steps).

We constructed 30 different Dutch stories to test participants’ reasoning abilities about other agents’ mental states. In the planned project, adult participants will listen to these stories with the help of headphones and watch the related drawings on the computer. They will be expected to answer a few questions about each of the stories (see Arslan, Verbrugge, Taatgen, & Hollebrandse for a study with children).

In this project, you will investigate adults’ eye-movements while they are answering questions in which they are supposed to reason about other agents’ mental states. You are expected to test 10 participants between the ages 18 and 35 to conduct the experiment, to analyse the eye-tracking data, and to write a report.

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl and Burcu Arslan, e-mail b.arslan@rug.nl

References

  • B. Arslan, N.A. Taatgen and R. Verbrugge, Modeling developmental transitions in reasoning about false beliefs of others. In: R. West and T. Stewart (eds.), Proceedings of the 12th International Conference on Cognitive Modelling. Ottawa: Carleton University, 2013, pp. 77-82. PDF
  • B. Arslan, R. Verbrugge, N. Taatgen and B. Hollebrandse, Teaching children to attribute second-order false beliefs: A training study with feedback. In: D. Noelle and R. Dale (eds.), Proceedings of the 37th Annual Conference of the Cognitive Science Society, Cognitive Science Society, Pasadena, CA, 2015 (to appear).

Second-order theory of mind reasoning in ACTransfer

The false belief task is a standard test for people’s theory of mind abilities. The second-order false belief task test whether people can also engage in second-order theory of mind, in which they reason about the way other people use theory of mind. In this project, we aim to model the process of second-order theory of mind in ACTransfer, an architecture designed to identify similarities across different tasks. In particular, we want to create a model of the second-order false belief task in ACTransfer.

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl and Burcu Arslan, e-mail b.arslan@rug.nl

References:

  • B. Arslan, N.A. Taatgen and R. Verbrugge, Modeling developmental transitions in reasoning about false beliefs of others. In: R. West and T. Stewart (eds.), Proceedings of the 12th International Conference on Cognitive Modelling. Ottawa: Carleton University, 2013, pp. 77-82. PDF
  • S. Wierda and B. Arslan, Modeling theory of mind in ACTransfer (extended abstract). In: J. Szymanik and R. Verbrugge (eds.), Proceedings of the Second Workshop Reasoning About Other Minds: Logical and Cognitive Perspectives, CEUR Workshop Proceedings, Volume 1208, 2014, pp. 40-44. PDF

Modeling theory of mind reasoning in games

Agent-based simulations suggest that theory of mind can be useful across a wide variety of games. In this project, we aim to model the use of theory of mind across these games in ACTransfer, an architecture designed to identify similarities across different tasks. Possible games include rock-paper-scissors, stag hunt, battle of the sexes, and limited bidding. Depending on your interests, this project can take on different forms. For example, the focus could lie more on modeling or on empirical study.

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl and Harmen de Weerd, e-mail h.a.de.weerd@rug.nl

References:

  • H. de Weerd, R. Verbrugge and B. Verheij, How much does it help to know what she knows you know? An agent-based simulation study. Artificial Intelligence, 199-200 (2013) 67-92. DOI:10.1016/j.artint.2013.05.004.

The role of feedback in the development of reasoning about situations like “Mary thinks that John thinks that there is a chocolate in the drawer”

Previous studies showed that reasoning about an agent who is reasoning about another agent develops between the ages of five and six (Sullivian et al., 1994). Arslan, Taatgen and Verbrugge constructed a computational cognitive model in ACT-R to explain the developmental transitions in reasoning about false beliefs of others (for the model see http://www.ai.rug.nl/~barslan/styled/index.html).

This project will test the predictions of the model with the 75 children around the age of six.

Children will listen to stories with the help of headphones and watch the related drawings on the computer. They will be expected to answer the related questions about the stories. In addition to the stories, children will be presented with a memory task and a computer game. Since it will be a training study, one child will be tested in 4 different days in order to track the effect of the training.

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl

Burcu Arslan, e-mail b.arslan@rug.nl

Bachelor student: Maximilian Seidler

Using intrinsic complexity of turn-taking games to predict people’s reasoning strategies

A recent study by Meijering et al. showed that when they play the game Marble Drop, people do not seem to be using the reasoning strategy that game theory predicts, called backward reasoning. Instead they appear to use a reasoning strategy that we call ‘forward reasoning plus backtracking’ (see for the paper: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0045961 ).

Szymanik, Meijering and Verbrugge are following up on this research. They have used a theoretical analysis of complexity to predict people’s reaction times in Marble Drop (see for the paper: http://www.jakubszymanik.com/papers/AlternationsMDGfinal.pdf )

In the bachelor project, you will perform data analysis on existing data about people’s reaction times and / or eye movements in the Marble Drop game in order to explain which reasoning strategies they may be using.

Supervision: Rineke Verbrugge, e-mail L.C.Verbrugge@rug.nl

Bachelor student: Gerben Bergwerff