-
What path(s) do high-performing students take through online
educational materials?
What social networks can foster or inhibit learning?
Do users of online learning tools behave as the system designers
expect?
What diagnostic substructures are commonly found in
studentproduced diagrams?
Can we use prior student data to identify students' solution
plan, if any?
Can we use prior student data to provide meaningful hints in
complex domains?
Can we identify students who are particularly helpful based
upon their social interactions?
Thus, graphs are simple in concept, general in structure, and
have wide applications for Educational Data Mining (EDM).
Despite the importance of graphs to data mining and data
analysis there exists no strong community of researchers focused on
Graph-Based Educational Data Mining. Such a community is
important to foster useful interactions, share tools and techniques,
and to explore common problems.
2. GEDM 2014
This is the second workshop on Graph-Based Educational Data
Mining. The first was held in conjunction with EDM 2014 in
London [
17
]. The focus of that workshop was on seeding an initial
community of researchers, and on identifying shared problems, and
avenues for research. The papers presented covered a range of
topics including unique visualizations [
13
], social capital in educational
networks [
8
], graph mining [
19, 11
], and tutor construction [
9
].
The group discussion sections at that workshop focused on the
distinct uses of graph data. Some of the work presented focused
on student-produced graphs as solution representations (e.g. [
14,
3
]) while others focused more on the use of graphs for large-scale
analysis to support instructors or administrators (e.g. [
18, 13
]).
These differing uses motivate different analytical techniques and,
as participants noted, change our underlying assumptions about
the graph structures in important ways.
3. GEDM 2015
Our goal in this second workshop was to build upon this nascent
community structure and to explore the following questions:
1. What common goals exist for graph analysis in EDM?
2. What shared resources such as tools and repositories are
required to support the community?
3. How do the structures of the graphs and the analytical methods
change with the applications?
The papers that we include here fall into four broad categories:
interaction, induction, assessment, and MOOCs.
Work by Poulovassilis et al. [
15
] and Lynch et al. [
12
] focuses
on analyzing user-system interactions in state based learning
environments. Poulovassilis et al. focuses on the analyses of
individual users' solution paths and presents a novel mechanism
to query solution paths and identify general solution strategies.
Lynch et al. by contrast, examined user-system interactions from
existing model-based tutors to examine the impact of specific
design decisions on student performance.
Price & Barnes [
16
] and Hicks et al. [
6
] focus on applying these
same analyses in the open-ended domain of programming. Unlike
more discrete tutoring domains where users enter single equations
or select actions, programming tutors allow users to make drastic
changes to their code on each step. This can pose challenges for
data-driven methods as the student states are frequently unique
and admit no easy single-step advice. Price and Barnes present a
novel method for addressing the data sparsity problem by focusing
on minimal-distance changes between users [
16
] while in related
work Hicks et al. focuses on the use of path weighting to select
actionable advice in a complex state space [
6
].
The goal in much of this work is to identify rules that can
be used to characterize good and poor interactions or good and
poor graphs. Xue at al. sought address this challenge in part via
the automatic induction of graph rules for student-produced
diagrams [
22
]. In their ongoing work they are applying evolutionary
computation to the induction of Augmented Graph Grammars,
a graph-based formalism for rules about graphs.
The work described by Leo-John et al. [
10
], Guerra [
5
] and
Weber & Vas [
21
], takes a different tack and focuses not on graphs
representing solutions or interactions but on relationships.
LeoJohn et al. present a novel approach for identifying closely-related
word problems via semantic networks. This work is designed to
support content developers and educators in examining a set of
questions and in giving appropriate assignments. Guerra takes
a similar approach to the assessment of users' conceptual changes
when learning programming. He argues that the conceptual
relationship graph affords a better mechanism for automatic
assessment than individual component models. This approach is
also taken up by Weber and Vas who present a toolkit for
graphbased self-assessment that is designed to bring these conceptual
structures under students' direct control.
And finally, Vigentini & Clayphan [
20
], and Brown et al. [
2
]
focus on the unique problems posed by MOOCs. Vigentini and
Clayphan present work on the use of graph-based metrics to
assess students' on-line behaviors. Brown et al., by contrast, focus
not on local behaviors but on social networks with the goal of
identifying stable sub-communities of users and of assessing the
impact of social relationships on users' class performance.