Research Statement

Dr. Oleksandra (Sasha) Poquet
January 21, 2020

My research agenda aims to understand how human, technological, and artificially intelligent agents affect peer learning behaviour and knowledge processes in the context of digital learning. I use computational methods to research social interactions using interdisciplinary analytical frameworks. My work is situated in the applied domain of learning analytics.

As an educational scientist, I am motivated by the ethos that educational systems need to educate learners holistically, rather than merely support their knowledge acquisition and upskilling. With the increased prominence of online and blended learning, instructors need to design richer interpersonal experiences for digital learning settings. To provide feedback about social learning online, I study digital social learning across 3 directions: 1) how multi-level networks of learners form; 2) how to model processes and outcomes of social learning; and 3) what mechanisms enable peer effects on learning behaviour via digitally-mediated communication. I view complexity science as a paradigm to capture learning processes across educational systems. I am exploring the use of complexity science 1) to capture and understand random, emergent, and ordered processes of learning through interpersonal interactions; and 2) model the dynamics of social learning.

Effective Learner Networks – Ongoing Work
Digitally-mediated learning environments are unique spaces that unfold from individual motivations, instructor prompts and curriculum design, and due to the emergent social processes. My ongoing interest is to understand topologies and formation of learner communication networks in digital settings. This understanding can help provide instructors with feedback about their design of social learning. In MOOCs for me this interest translates into community identification and formation mechanisms. I view learner interactions in MOOCs as multi-level networks of committed learners and intermittent visitors. This framing enabled me to disentangle the hierarchy between multi-level network actors. My research showed how they differ in social presence perceptions [1], conversation dynamics and discussion topics [2], and accumulation of social capital [3]. Using these distinctions, I proposed community indicators [4], showed how these indicators relate to forum facilitation styles [5], and replicated the indicators in another technological setting [6].

To provide feedback on the design of social learning in university online courses, I study the relationship between learning design and forum networks. I conceptualise learner interactions as multi-level networks of posting activity, communication structure, and social relations. Using this framework enabled me to observe the link between network topologies with pedagogical tasks. By reducing heterogeneity of posting behaviour to individual and networked orientation, I reveal underlying archetypal structures of communication and social networks. I found that small group activities generate small-world-like topologies, tasks that prompt networked activity produce random graphs, and tasks that promote individual-level reflection generate preferential attachment networks [7].

My future interest in this area is to apply spatial and temporal modelling to understand the meaning and role of space and time in formation of online networks [8]. Another research interest is modelling relations across multi-level network of online communication spaces.

Measures of Social Behaviour – Ongoing Work
Besides the topology and formation of learner networks, I am interested in how individual-level behaviour describes the process and outcomes of social interactions in online discussions. Assessing individual-level behaviours can support learners in improving their competences. Simple measures of social learning behaviour are also instrumental to the formation of macro-scale patterns. Interventions towards better social learning are contingent on how well this micro-macro relationship is understood.

To provide learners with feedback on their behaviour, my research investigated the meaning of network centrality in online forums. I found that forum network positioning co-evolves from random co-occurence to social selection of conversation partners [9]. As learners co-occur indirectly in discussion affiliation networks, they are more likely to interact directly in the future. With a group of collaborators, we showed that learner centrality in online forums is a function of individual activity and skill; weighted local clustering may reflect emergent social processes instead [10]. Through a long-standing collaboration, I rely on NLP to understand network positioning. This work demonstrated that network positioning is underpinned by distinct linguistic properties of narrativity and cohesion [11]. We further linked micro-level communication characteristics (e.g. newness, social impact, responsivity, participation, communication density) with measures of learner interpersonal positioning that span across these micro-contexts [12].

Future work in this area includes dynamic modelling to explore if individual communication behaviour at micro time scale provides semantic space for self-organization, whereas relational centrality captures self-organisation outcomes observable at larger time-scales. Another direction differentiating students through the recurrence analysis of communication patterns as a window to their relational positioning. I am interested in how cell behaviour and evolution hypotheses can be used in researching social learning online.

Peer Effects – Current Work
My ultimate research interest is to facilitate the contagion of learning behaviour through digital learner communication. Evidence of peer effects in digital settings, particularly learning behaviour adoption through exposure, has been limited. To this end, I studied longitudinal data from 4-years of student interactions from the National University of Singapore, University of South Australia, and the University of Michigan. Analysis supports the presence of peer effects on performance and social selection at high-performer level [13]. This early work carries a number of limitations. I have collected several datasets where I am exploring contagion of various learning outcomes as well as process of linguistic entrainment as proxy for contagion. Given the specificity of the digital communication data, I am also developing a repertoire of modelling techniques more suited to the sparse asynchronous communication data.

To extend the impact of my work beyond empirical insights towards the feedback to learners and instructors, I collaborate with educational practitioners and educational data scientists who serve university stakeholders. I am working towards securing research funding that enables collection of high quality online and face-to-face learning interaction data in an educational context that caters to non-traditional and first-generation students. I intend to apply my methodological and analytical approaches to understand social processes, learning outcomes, and peer effects to design interventions that help develop their lifelong learning skills.

[1] Poquet et al. (2018). Social Presence in Massive Open Online Courses.
[2] Poquet et al. (Under review). Communication patterns of residents and visitors in MOOCs.
[3] Joksimovic, Dowell, Poquet, et al. (2018). Exploring the development of social capital in cMOOC through discourse
[4] Poquet & Dawson (2018). Network patterns of direct and indirect reciprocity in edX MOOC forums.
[5] Poquet et al. (2017) How effective is your facilitation?
[6] Poquet et al. (2018). Are MOOC forums changing?
[7] Poquet et al. (Under review). Multi-level approach to online forum evaluation.
[8] Chen & Poquet (2020). Socio-temporal events in peer interactions.
[9] Poquet & Jovanovic (2020). Interpersonal and intergroup positioning in online forums
[10] Poquet et al. (2020). Are forum networks social networks?
[11] Dowell, Poquet, et al. (2015). Modeling learners’ social centrality and performance through language
[12] Dowell & Poquet (Under review) Identifying Learner Roles through Group Communication and Interpersonal Network Positioning
[13] Poquet et al. (In preparation). Homophily or social selection in higher performers ties online.