Exploring Classroom Discourse Through an Agent-Based Model

Elizabeth B. Dyer, Northwestern University, elizabethdyer@u.northwestern.edu

Introduction Classroom discourse is seen as a key component for supporting student learning [2]. However, supporting substantive student discourse about mathematics is very difficult for teachers [1, 3]. This poster describes the design of an agent-based model of debriefing discussions to help teachers explore the relationship between individual interactions and classroom discourse. Agent-Based Modeling •  Computational representation for

simulating complex systems [5] •  Behaviors and interactions of

individuals (agents) in a system are coded (the rules)

•  NetLogo is an agent-based coding environment (http://ccl.northwestern.edu/netlogo/)

Discourse as an Agent-Based Model •  Interactions between agents

(students and teacher) produce discourse at the classroom level

•  Discourse becomes an emergent phenomena [4]

Student Actions Express Idea – Share an idea Express Disagreement – Disagree with another student’s idea Express Agreement – Agree with another student’s idea Ask Clarifying Question – Ask another student to clarify unclear ideas Student Characteristics Idea – What idea the student has Clarity – Level of clarity expressing that idea Participation – Likelihood of participating in discussion Responding – Likelihood of responding to other students’ ideas Questioning – Likelihood of asking clarifying questions to other students when responding

Graphs Showing Discourse Aspects Teacher talk – frequency of different teacher actions Types of dialog – frequency of agents involved in responsive dialog Changing idea – frequency of students changing their ideas Student ideas – histogram of the number of students with each idea Participation – distribution of participation level in students

Open Model Exploration with Guiding Questions

•  What combination of teacher moves produce the most student-to-student talk?

•  What combination of teacher moves produce a high percentage of students with the correct idea in few turns?

•  What teacher actions are most likely to change students’ ideas?

•  Do these combinations depend on student parameter levels?

Teacher Actions Ask Open-ended Questions – Ask class to answer a question Clarifying Questions – Ask a student to clarify idea Advancing Questions – Ask questions that challenge or further a student’s idea Questioning Specific Students – Ask specific students to share idea Encourage Responding – Ask if students have any questions or comments for the current speaker Evaluate – Evaluate student ideas as correct or incorrect Explain – Give explanations of mathematics ideas

Literature Informing Model Design Boaler, J., & Brodie, K. (2004). The importance, nature and impact of teacher questions. Proceedings of the 26th annual meeting of

the North American chapter of the International Group for the Psychology of Mathematics Education (Vol. 2, pp. 773–781). Toronto, Ontario, Canada.

Hufferd-Ackles, K., Fuson, K. C., & Sherin, M. G. (2004). Describing Levels and Components of a Math-Talk Learning Community. Journal for Research in Mathematics Education, 35(2), 81–116.

Smith, M. S., & Stein, M. K. (2011). Five Practices for Orchestrating Productive Mathematics Discussions. Reston, VA: National Council of Teachers of Mathematics.

References [1] Hufferd-Ackles, K., Fuson, K. C., & Sherin, M. G. (2004). Describing Levels and Components of a Math-Talk Learning Community. Journal for Research in Mathematics

Education, 35(2), 81–116. [2] National Council of Teachers of Mathematics. (2000). Principles and Standards for School Mathematics. Reston, VA: Author. [3] Sherin, M. G. (2002). A balancing act: Developing a discourse community in a mathematics classroom. Journal of Mathematics Teacher Education, 5(3), 205–233. [4] Wilensky, U. (2001). Modeling nature’s emergent patterns with multi-agent languages. Proceedings of EuroLogo 2001. Linz, Austria. [5] Wilensky, U., & Resnick, M. (1999). Thinking in levels: A dynamic systems approach to making sense of the world. Journal of Science Education and Technology, 8(1), 3–19.

Transcript Excerpt from Model T: Which recipe will be more lemony? S 10: It will be the same lemony. T: You're wrong! T: Which recipe will be more lemony? S 28: The 2nd recipe is more lemony because it has 1 more lemons. T: The 2nd recipe does have more lemons, but it also has more waters. T: Which recipe will be more lemony? S 25: The 2nd recipe is more lemony because it has 3 lemons, but the 1st

recipe has 2 lemons, so it has 1 more lemon. T: You're wrong! T: Which recipe will be more lemony? S 16: The 2nd recipe is more lemony because it has 3 lemons, but the 1st

recipe has 2 lemons, so it has 1 more lemon. T: The 2nd recipe does have more lemons, but it also has more waters.

Transcript Excerpt from Model T: Which recipe will be more lemony? S 16: It will be more lemony. T: What did you mean? S 16: There is a higher percentage of lemons. T: Does anyone have a question or reply to Student 16's comment? S 6: I disagree. The 2nd recipe is less lemony because it has 1 more water

than the first recipe. T: Does it matter how many lemons the recipes have? S 6: I change my mind. The 2nd recipe is more lemony because it has a

higher percentage of lemons. T: Interesting idea. T: Which recipe will be more lemony? S 22: Both recipes are equally lemony because you can make the 2nd recipe

by adding 1 water and 1 lemon to the 1st recipe. T: Does anyone have a question or reply to Student 22's comment? S 29: I agree. It will be the same lemony.

The model shows a classroom with a number of students and a teacher, with the dialog at the top. Turn by turn, someone adds a piece of dialog to the discussion. What and who talks is based on what was previously said, along with the student and teacher parameters set at the side of the model. Two Modes of Use 1.  Student and Teacher parameters set by user (default) 2.  Only student parameters set; user makes decisions for

teacher agent

Analysis of Classroom Discourse 1.  Model typical discourse

•  What parameter settings produce the typical discourse in your classroom?

•  Are you surprised by the settings that produce typical discourse?

2.  Model changes to typical discourse •  How does slightly changing the teacher

parameters in your typical model change discourse?

•  Which of these changes would improve your discourse?

Potential Use with Teachers

Initiation-Response-Evaluate Discourse Example

Teacher-Supported Productive Discourse Example

Design of Model