Movie

1
Movie 2 -- 10th Grade BiologyFrom loose
recitation to free time
2
Interaction Orders, Network Switches, and the
Process of Schooling

• Daniel McFarland
• mcfarland_at_stanford.edu
• Skye Bender-deMoll
• skyebend_at_stanford.edu

Research presented here is supported by funds
from the Spencer Foundation and Stanford
Universitys Office of Technology and Licensing.
3

• Topic / Problem -- two interrelated goals of
  paper.
• The first is to explain how interaction dynamics
  are central to our understanding of the structure
  of complex organizations.
• Interaction dynamics reveal how normative
  structures of association form in everyday
  affairs how they entail a multiplicity of
  coordinated endeavors, some of which involve
  orthogonal forms of association how actors can
  move between this multiplicity of affairs and
  how local actions can depart from and transform
  larger aggregate patterns of association.
• This addresses a shortcoming in network theory.
  The least developed aspect of social network
  analysis concerns the fluid, interpretive
  dynamics of communication on the micro-level.
  Fluid interactions are generally assumed to
  aggregate into stable relations and cultural
  domains are either fully ignored, or only related
  to networks in a casual, ad-hoc basis.

4

• Topic / Problem -- two interrelated goals
  (continued)
• The second goal is to apply this
  conceptualization to the empirical dimension of
  schools.
• We illustrate how interaction dynamics afford a
  deeper understanding of schooling. Interaction
  dynamics reveal how a teacher-centered pattern of
  pedagogy arises in most classrooms how pedagogy
  nonetheless entails a variety of coordinated
  endeavors that depart from this norm how
  participants move through these different and
  orthogonal forms of coordinated activity and how
  local actions can induce departures,
  negotiations, and transformations of the larger
  normative structure of schooling.
• The process of classroom order is important for
  education research because it is a precursor to
  learning (content focus). As patterns of
  interaction settle into recognized patterns,
  teachers and students find opportunity to shift
  attention away from concerns of coordination and
  social order to subject matter content. Current
  education research does not study the
  distributed, relational nature of educational
  settings in a systematic way that can be
  formalized and tested. This work presents such
  analysis.

5

• Theoretical Argument on Structuring Process
• The structuring of classrooms entails at least
  four classes of interaction order role-frames,
  transitions, bracketed laminations, and problems
  (White 1995 Mische and White 1998). Certain
  classes of interaction order entail concatenating
  actions that reinforce norms, while others
  facilitate switches and counter-normative
  behaviors.
• Role-frames entail interaction where networks and
  interpretive frames concatenate and reinforce
  each other so as to create specialized forms of
  stable interaction. In the classroom, there is
  an academic role-frame and a social role-frame
  anchored in certain topics, activities, social
  identities, and status structures.
• Transitions arise as unstructured segments at the
  beginning and end of class, as well as between
  tasks when the academic frame and social roles of
  friendship and clique membership are not on.
• Problems are volatile episodes when the social
  order breaks down, uncertainty takes hold, and
  then reconstruction or repair efforts follow. As
  such, they are interchanges that can emerge in
  the middle of other classes of interaction order
  and sweep aside their dynamics.
• Another type of interaction order consists of
  bracketed laminations that span interpretive
  frames by bracketing episodes of interaction
  wherein a role-frames meaning is transformed
  into something patterned on but independent of
  them (jokes, keying, fabrications, etc).
• Building on prior work, we describe interaction
  ordering, or structuration, as guided by rules
  (activity-scripts) and resources (Giddens 1984
  Barley 1986 Sewell 1992), but we expand the
  notion of resources to entail (a) claims
  (discursive cues) and (b) contexts (network
  configurations).

6

• Empirical Argument on the Structure of Schooling
• We reinforce the conceptual argument through case
  studies and formal models of classroom
  interaction. In particular, we identify processes
  that establish (or undermine) recursive patterns
  of behavior.
• We find that there are certain activities,
  network conditions, and forms of talk that
  generate the normative structure of schooling
  across classrooms.
• We find the normative structure of schooling is
  teacher-centered (e.g., recitation and lectures),
  and that even in decentralized activities of
  class discussion, there is a tendency to fall
  back on centralized formats to coordinate
  collective action.
• In addition, we identify (de)stabilizing
  mechanisms within types of activities (i.e.,
  coordination games), which reveal how the
  structure of schooling can be transformed by
  practitioners.
• Only in student-centered tasks like presentations
  and group work does the teachers diminished role
  help stabilize the pattern of ongoing
  interactions, and reveal routes by which
  alternative forms of schooling can be
  successfully managed.

7
Data and Unit of Time

• Unique Data on Streaming Classroom Interactions
• Observed many classrooms and recorded who spoke
  to whom, when, and in what manner.
• Caught turns of interaction and their
  corresponding forms of speech (task, social,
  moral evaluations, technical evaluations,
  broadcasts, questions), and noted activity
  schemas, time, etc.
• Lower reliability is had in phases where
  simultaneous talk arises (free time). However,
  all measures use patterns of interaction (instead
  of frequency), and the number of observations
  decreases the amount of error affecting results.
  In spite of reliability issues, this is closest
  data has come to coded streaming interactions,
  and it is useful for developing methods
  appropriate to data on fully dynamic systems.
  Should there be a bias, it will underreports of
  such behavior.
• The final data set consists of over 800,000 turns
  of interaction that spans 3000 individuals, 603
  class periods, and 153 separate classrooms of two
  high schools.
• Classroom focus
• We study how classrooms organize interaction over
  time.
• Our goal is to identify mechanisms endogenous to
  classroom settings and to offer a bottom-up
  perspective that is remiss in most sociological
  and educational research.
• How do we decide on the unit of time to use?
• Since we are interested in how interaction is
  ordered, we want a time-scale that is not too
  short (so as to miss patterning of different
  games) nor too long (so as to merge multiple
  doings together). Hence, we settle on 2.5 minute
  slices of time, aggregate within them, and then
  look at how slices of class time are
  interrelated.
• Other durations of time could be used and the
  exploration of this is the topic of a future
  paper (Sunbelt topic).

8

• Measurement When does order arise in
  interaction?
• Challenge to measure stability when interaction
  patterns vary depending on the activity being
  performed. We want a single measure of
  stabilization.
• How do we capture the stability in interaction
  patterns when ties are distributed and can have
  variable organization in different settings? We
  take the triad census for every slice of class
  time and correlate it with the successive slice
  (census of slice t0 with census of slice t1).
• Since null triads are overwhelmingly common, we
  omit them
• Also, we separate task from sociable types of
  ties ( 30 triad types).
• The correlation of successive triad censuses is a
  measure of pattern-similarity over time, and
  affords a scaled variable that is consistent
  across classes regardless of the type of activity
  and the size of the classroom network.
• We tried other distance measures that had various
  problems hamming distances, euclidean distances
  (across dyads and linguistic arrays), and even
  clustered structural distances (Butts and Carley
  2003).

9

• RESULTS (cases and formal models)
• Part 1 -- Four cases of interaction order
• Will use variety of visualizations to reveal
  dynamic processes of network manipulation and
  task cueing.
• Summary plots
• Used to display hundreds of thousands of
  observations in a format that can be readily
  understood and explored by practitioners and
  researchers (rates and inter-structural
  correlations are plotted).
• Visualization of network dynamics
• SoNIA (Social Network Image Animator,
  Bender-deMoll and McFarland 2003 Moody,
  McFarland and Bender deMoll 2004) animates
  across graph layouts and uses a series of
  techniques to render movement reliable and
  meaningful. Free software publicly available
  lthttp//sonia.stanford.edugt

10

• Two Cases of Switching
• Case 1 Trigonometry
• announcements, recitation review, recitation fun
  problem, pair work quizzes
• Case 2 Biology announcements, recitation lesson
  (with humor), recitation review, free time

11
(No Transcript)
12

• Movie 1 Trigonometry announcements,
  recitation review, recitation fun problem, pair
  work quizzes
• Movie 2 Biology announcements, recitation
  lesson (with humor), recitation review, free time

13

• Two Cases of Stalled /Failing Order
• Case 3 Economics
• Lecture introduction, groupwork
• Case 4 English 10
• Lecture introduction, seatwork

14
(No Transcript)
15

• Movie 3 Economics Lecture introduction,
  groupwork
• Movie 4 English 10, Speech Lecture
  introduction, seatwork

16

• Finds across four cases
• Different games with different organization.
• Different dances that designate distinct
  interaction orders
• recitation entails stars with rotating core
• free time has fluid cluster and spanning tree
• group work has stable cliques and teacher
  monitoring
• student presentation entail rotating stars
• seatwork entails local, fluid communications
  centered around teachers
• Different network processes
• Centralization-decentralization (e.g.,
  recitation/presentation versus group work / free
  time)
• Clustering-separating (e.g., group work versus
  seatwork)
• Different points of disorder
• brackets of period
• cued transitions
• unplanned deviations and rebellion
• Typically inverse network processes as that
  called for by the activity. For example, group
  work falls apart with broadcasts or
  centralization moves, and lectures fall apart
  when clustering and/or competing centralization
  efforts arise.

17

• Part 2 Formal models
• Additive model of central tendency
• Want to identify general structure of schooling.
• Correlation is really skewed variable with
  classes veering into and out of interactional
  stability -- basically a discrete shift. We
  dichotomize the variable so that stable refers
  to slices of class time with triad patterns
  correlated by at least .89 (value is the median
  correlation).
• We then use fixed effects models that identify
  mechanisms creating stability/change within
  classroom interaction over time (so classroom and
  school level characteristics are removed).
• We use fixed effects because it enables us to
  focus on the social processes endogenous to
  classrooms (where most of the variance resides)
  and it affords the most conservative model (i.e.,
  removes spuriousness, bias, heterogeneity
  shrinkage, etc).
• Interaction terms
• Conditional effects presented since we want to
  know how various network conditions and forms of
  speech influence the stability of interaction
  within types of prescribed tasks (Jacard 2003).
• We want to know the relative rate of each form of
  speech relative to other forms. We operationalize
  this as the percentage of total interaction that
  is of a particular type. Such measures capture
  the stylistic focus within a slice of class time
  as opposed to sheer volume or rates of behavior.
  We argue that it is linguistic style that should
  affect interaction roles and the structuring
  process.

18
(No Transcript)
19

• Factors causing interaction order/disorder in
  classes story of centralization and control,
  but more from local disputes and coordination
  problems (bottom up perspective of schooling
  process).
• Transitions are more volatile (open-close,
  transitions)
• Long-term decrease in stability (months into
  year)
• Cued activity structures differ in order /
  disorder
• lectures gt presentation, recitation, discussion gt
  group work, tests, audio gt seatwork and
  undefined.
• Centralized formats tend to be more stable
  routines (lacks ambiguity of hierarchy).
• Forms of Speech central tendencies of schooling
  process
• (driven by most common activities, i.e.
  recitation)
• Teacher broadcasts and technically evaluative
  speech stabilizes
• Most any form of direct talk destabilizes, but
  especially teacher-student talk, conflicts
  (moral) and bracketed laminations (jokes).
• Network conditions vary in effect
• Task asymmetrical ties are helpful (broadcasts),
  social reciprocity creates some disturbance, and
  reciprocal task ties and asymmetric socializing
  are quite problematic (student outbursts).

20

• Prior section describes central tendencies of
  schooling process, and as such, describes the
  realized social structure emerging in classes
  more generally. The emergent structure of
  schooling is contingent on which activities are
  prescribed and enacted.
• The tendency toward centralized formats reflects
  problems of coordination in student-centered and
  progressive forms of instruction.
• In order to transform the structure of schooling,
  the problems of coordination undermining
  progressive forms of instruction need to be
  understood and overcome. Therefore, it is useful
  to know what types of behavior sustain or
  undermine distinct interaction orders. In effect,
  we ask, what speech and network forms are crucial
  to the stability of different activity
  structures, and which are most capable of
  undermining them. We turn to interaction terms

21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24

• Lecture, recitation, and discussion all similar
  in that they are teacher centered
• Lecture best maintained when the teacher voices
  broadcasts and technical evaluations undermined
  when the teacher is sociable, or when students
  have side conversations, argue, and emit
  broadcast statements of their own.
• Recitation same as lecture but now student
  technical evaluations also help stabilize
  interaction and T-S contact is more acceptable
  undermined by teacher-sociability, student
  broadcasts, and peer side conversations.
• Discussion best maintained by teacher broadcasts
  and questions, and there is tolerance for more
  open dialogue undermined by teacher-student
  sociability, jokes, and student arguing.
• Student presentations different
• Presentation student broadcasts stabilize, but
  so do teacher broadcasts/questions, and peer task
  interaction (Tch MC who coordinates and
  discusses / embellishes student broadcasts)
  undermined by open socializing and arguing.

25

• Group work, undefined, seatwork, and tests /
  audiovisuals
• Group work best maintained by a distant teacher
  and student task conversations tolerant of
  sociable behavior and humor, but undermined by
  any centralizing claims (broadcasts).
• Undefined (free time / maintenance) maintained
  by student socializing undermined by peer task
  interactions and teacher broadcasts.
• Seatwork tolerant of peer socializing, humor,
  and students posing questions undermined by
  centralized formats.
• Tests teacher questions and student technical
  evaluations mildly assist matters (i.e., test
  preparation behaviors), but everything else
  disrupts the activity.
• Audiovisual non-action, or silence, stabilizes
  the task.

26

• Network Conditions and Cued Tasks
• Task asymmetry helpful for centralized tasks
  like lecture, recitation, discussion (teacher
  centered discussions Socratic method), and
  student presentations.
• Task reciprocity key for group work and not much
  else.
• Social asymmetry avoid in most tasks, but not of
  concern in undefined segments. Entails student
  broadcasts, so especially disruptive of routine
  segments involving repetition (lecture,
  recitation, group work) or low interaction rates
  (tests and audiovisuals) where it stands out.
• Social reciprocity salient in seatwork and
  undefined segments where appropriate, but
  problematic elsewhere.

27

• Conclusion
• Study identifies general structure of schooling
  as well as the strengths and weaknesses of
  various instructional formats.
• Models show the dynamics of interaction, and how
  these dynamics are organized in ways reflective
  of a general schooling process.
• The general story derived from these results is
  that the structure of schooling is defined
  primarily by prescribed activity-frames and
  issues of instructional control. Activity frames
  define the instructional process and the network
  conditions educators want to establish. However,
  prescribed formats do not always take hold, and
  interaction orders of varying kinds can drift
  toward more centralized formats.
• Results offer some hope in that educators can
  identify various activities, network conditions,
  and forms of speech (i.e., rules, contexts, and
  claims) that can either bolster or transform
  interaction patterns they would like to establish
  in educational settings.

28
Interaction Orders, Network Switches, and the
Process of Schooling

• Daniel McFarland
• mcfarland_at_stanford.edu
• Skye Bender-deMoll
• skyebend_at_stanford.edu

Research presented here is supported by funds
from the Spencer Foundation and Stanfords Office
of Technology and Licensing.
29

• Histogram for Triad Census Correlation
• (Mean .6639, Std Deviation .3866, Median .8887)
• Histogram
  Boxplot
• 1.025
  602
• .
  3219 -----
• .
  918
• .
  458 -----
• .
  251
• .
  176
• .
  193
• .
  387
• .
  281
• .
  229
• .
  181
• .
  156
• .
  139

30

• Group dramatization in creative writing classs

31
(No Transcript)