Example of research using agent-based modeling methodology to investigate individual and
social factors underlying inequitable participation patterns observed in a real classroom (environment) in
which an experimental collaborative activity. Researchers created agent-based
simulations of simplified collaborative activities and qualitatively compared results from
running the model with the classroom data. They found that collaboration pedagogy
emphasizing group performance may forsake individual learning, due to preference for short term
group efficacy over individual long-term learning.
A classroom engaged in collaborative group work can be seen as a complex adaptive system (Hurford,
2004) in which optimal as well as sub optimal behavioral patterns may emerge. Despite individual students’ initially exploratory behaviors, once a functioning coordination scheme evolves in a group and is evaluated as
well adapted to performing the mandated task, an implicit quietus is set on any further exploration or task
rotation, and the group achieves dynamic stability. Such arrangement would be fitting for workplaces, but its instantiation in classrooms may present teachers with the dilemma of maximizing group production at the expense of individual learning, especially of struggling students who are benignly assigned by the group to mathematically lesser tasks. ABM methodology may provide education researchers and practitioners tools for understanding such classroom dynamics, so that they can identify points of leverage for working with students’ natural behavioral inclinations to achieve equitable participation.
we chose a simpler numeric puzzle task (see Figure 1a).
This linear puzzle consists of set of numbered pieces to be concatenated in ascending order (1, 2, 3, 4…).
Necessary activities within this task are retrieving pieces (simplest task), verifying if pieces are already present
in the puzzle (intermediate demand), and connecting pieces (most demanding task). Initially, puzzle pieces are
scattered all over the classroom. Piece-retrievers wander around and, when they find a piece, grab it and go back
to their group’s table, delivering it to the piece-verifier. The piece-verifier evaluates whether a copy of the piece
is already present (the puzzle cannot have repeated pieces). If so, the piece-verifier orders the piece-retriever to
discard the piece and bring a new one. If the piece is suitable, the piece-verifier delivers it to the piece connector, who will check if the piece fits the puzzle in its current state, and connect it to the puzzle. For each successful micro-task, students receive positive feedback in the form of an increment in their skill (speed and/or accuracy). Overall group performance is evaluated by the time-to-completion divided by the number of correct pieces. Our independent variables are: (a) pedagogical style (with or without mandated role rotation); (b)students’ initial skill level for each task and distribution of skill levels within students; (c) task difficulty.
after many sets of experiments over a large initial parameters set, we were able to plausibly demonstrate
relations between pedagogical practice and student learning, as follows: (i) The overall performance of groups
with mandated role-rotation decreased by approximately 40%; (ii) When student–agents were reinforced for
group performance rather than individual learning, students became entrenched within skills reflecting their initial skill-level distribution; However, when role rotation was mandated, even though production slowed down, more learning occurred, per student.
Increasing a low-level task skill (i.e.,increasing the number of puzzle pieces a retriever–student can bring to the group per time tick) appears to decrease the correct/incorrect puzzle-pieces ratio (failure ratio) linearly,whereas increasing the high-level task skill effects a non-linear trend .