Interactive Communication

The PSLC Interactive Communication cluster

Abstract

The studies in the Interactive Communication deal primarily with learning environments where there are two interacting, communicating agents, one of which is the student. The other agent is typically a second student, a human tutor or a tutoring system. They communicate, either in a natural language or a formal language, such as mathematical expression or menus. We are trying to find out why such instructional, dyadic, interactive communication is sometimes highly effective and sometimes less effective. Sometimes we study highly constrained forms of communication in order to vary isolated aspects, and sometimes we compare whole forms of communciation. Our hypothesis is simply that interactive communication is effective if it guides students to attend to the right knowledge components.

Background and Significance

Although instructional dialogue has been studied in classrooms (e.g., Lave & Wenger, 1991; Leinhardt, 1990) and workplaces (e.g., Hutchins, 1995; Nunes, Schliemann & Carraher, 1993), we are focusing on more tractable albeit still complex situations: dyadic instructional dialogues, namely dialogues between: (a) a human tutor and a human student, (b) two human students, or (c) A computer tutor and a human student. Moreover, the dialogue are task-oriented (Grosz & Sidner, 1986) in that the participants are working together on a task rather than simply conversing with no shared goals or with opposing goals.

Early studies focused on the structure of dyadic instructional dialogue (e.g., Fox, 1993; Graesser, Person & Magliano, 1995; MacArthur, Stasz, & Zmuidzinas, 1990). When later studies compared the learning that occurred during dialogue vs. less interactive instruction (e.g., VanLehn, Graesser et al., 2007; Katz, Connelly & Allbritton, 2003; Evens & Michael, 2006; Cohen, Kulik & Kulik, 1982), they found surprisingly mixed results. Only 60% of the studies showed that interactive communication caused larger learning gains than less interactive instruction.

The interactive communication cluster is undertaking the next step in this important line of research by investigating when different types of interactive communication are effective and why. Sometimes we compare highly constrained forms of communication in order to vary isolated aspects, and sometimes we compare constrained interactive communication to passive communication (e.g., reading).

Glossary

See Interactive Communication Glossary

Research question

What properties of interactive communication promote robust learning?

Independent variables

The independent variables (also called Treatment Variables) of the IC cluster appear as column headers in the matrix above. They are listed here with links to their glossary entries.

• Collaboration

• Vicarious learning

• Collaboration scripts

• Deep/Reflection questions

• Instructional explanation

• Prompted Self-explanation

• Tutoring feedback

• Error correction support

Dependent variables

Measures of normal and robust learning.

Hypothesis

Our central hypothesis is just a special case of the Knowledge component hypothesis: interactive communication is effective if it guides students to attend to the right knowledge components. The key words here are “guide” and “attend” because they may oppose each other. A dialogue that strongly guides the student may also cause the student to disengage and thus not attend to the knowledge component even if the student’s dialogue partner mentions them. On the other hand, an unguided dialogue may increase the student’s engagement but may skirt around the right knowledge components. That is, the assistance dilemma surfaces as the degree of learner control (a term from the older educational literature) or student initiative (a nearly synonymous term from the natural language dialogue literature).

Explanation

If we view a short episode of interactive communication as a learning event space, there could be three reasons why one treatment might be more effective than another:

(1) The learning event spaces might have different paths with different content. For instance, if one person contributes critical information that the other person lacks, then their joint learning event space has paths that are absent in the learning event space of the second person if that person were working solo. That is, the topology of one space might be better than the topology of the other.

(2) If the learning event spaces in the two conditions are the same, then the interactive communication treatment might cause the students to traverse different paths than the control students. That is, the path choices of one treatment might be better than the path choices of the other.

(3) If the learning event spaces are the same and the students take the same paths, they still might learn more in one condition than another because of the way that they traversed the path. For instance, having a partner observe the student as the student traverse a path might cause the student to be more attentive to details and to remember more. That is, the path effects might differ in the treatment vs. the control.

Descendents

Collaboration

When and how does collaboration between peers can increase robust learning? Problem solving, example studying and many other activities can be done alone, in pairs, or in pairs with various kinds of assistance, such as collaboration scripts. From the standpoint of an individual learner, having a partner offers more assistance than working alone, and having a partner plus other scaffolding offer even more assistance. Thus, the Assistance Hypothesis predicts an interaction between various forms of peer collaboration and students' prior competence.

• Learning from Problem Solving while Observing Worked Examples (Craig Gadgil, & Chi)

• The effects of elaborative dialog on problem solving and learning (Hausmann & Chi, 2005)

• The effects of interaction on robust learning (Hausmann & VanLehn, 2007)

• Collaborative Extensions to the Cognitive Tutor Algebra: Scripted Collaborative Problem Solving (Rummel, Diziol, McLaren, & Spada)

• Collaborative Extensions to the Cognitive Tutor Algebra: A Peer Tutoring Addition (Walker, McLaren, Koedinger, & Rummel)

• Supporting Conceptual Learning in Chemistry through Collaboration Scripts and Adaptive, Online Support (McLaren, Rummel, Harrer, Spada, & Pinkwart)

Questioning

When and how can asking the student questions increase the student's robust learning? What kinds of questions are best?

• Deep-level questions during example studying (Craig & Chi)

• Post-practice reflection (Katz & Connelly)

• Reflective Dialogues (Katz, Connelly & Treacy)

• Extending Reflective Dialogue Support (Katz & Connelly)

• Self-explanation: Meta-cognitive vs. justification prompts (Hausmann, van de Sande, Gershman, & VanLehn, 2008)

Tell vs. elicit

When a tutor knows that something needs to be said, she or he must decide whether to tell it to the tutee, try to elicit it from the tutee via a question or prompt, or just wait and hope that the tutee says it. Similarly, if a tutor knows that something needs to be done, the tutor can do it, elicit the action from the student or just wait. An instructional designer faces the same choices. For each thing that needs to be said or done in the instructional dialogue, should the tutor or the student be made responsible for it? For instance, should the tutoring system point out errors to the students or should the students detect their errors? In general, assistance is higher when the tutor does a portion of the instructional activity than when the student does it.

• Does it matter who generates the explanations? (Hausmann & VanLehn, 2006)

• Does Treating Student Uncertainty as a Learning Impasse Improve Learning in Spoken Dialogue Tutoring? (Forbes-Riley & Litman)

• Tutoring a meta-cognitive skill: Help-seeking (Roll, Aleven & McLaren) [Also relevant to Refinement & Fluency, Knowledge component analysis]

• Understanding culture from film (Ogan, Aleven & Jones) [Also relevant to Refinement & Fluency, Explicit instruction and manipulations of attention & discrimination]

• The self-correction of speech errors (McCormick, O’Neill & Siskin)

• Using Elaborated Explanations to Support Geometry Learning (Aleven & Butcher)