Reconnecting relationships through technology: Developing feedback literacy capabilities through an AI automated feedback tool

Synthesis: Tubino and Adachi (2022) reframe AI automated feedback tools as a vehicle for developing students' feedback literacy, not merely for reducing teacher workload. Drawing on a Deakin University-wide T&L pilot (2021) with FeedbackFruits' AI automated feedback tool across 29 units and nearly 4,000 students, they show the tool positions feedback as a student-centered activity and affords the demonstration of several dimensions of Molloy et al.'s (2020) learning-centered framework for feedback literacy. They propose making the tool student-facing and degree-wide so students decide what and when to seek feedback on, building agency and self-regulated feedback literacy without adding teacher workload.

Core Finding

What the Paper Does

The paper reports a University-wide teaching-and-learning pilot at a large Australian university in partnership with FeedbackFruits (a Netherlands-based edtech company) through a 'DoTank' co-design innovation project. The AI automated feedback tool is a form of automated writing evaluation (AWE) focusing on micro-level text features (sentence length, punctuation, grammar, text structure), supporting the copy-editing stage of writing: the teacher sets parameters while the student uses the tool independently, receiving timely, actionable feedback without teacher involvement. The authors use the seven-dimension feedback literacy framework of Molloy et al. (2020) — committing to feedback as improvement, appreciating it as an active process, eliciting information, processing feedback information, working with emotions, acknowledging reciprocity, and enacting outcomes into learning goals — as the lens for the tool's iterative, design-based (Reeves, 2006; van den Akker, 1999) design and evaluation.

Pilot Scale and Usage

• Ran across three teaching periods on 29 units, reaching almost 4,000 students across undergraduate and postgraduate levels in SEBE, Arts and Education, and Business and Law faculties; 70% of students with access were undergraduates.
• Writing tasks were diverse: theses, research and project reports, autoethnographies, essays, and reflective tasks.
• Usage averaged ~13% of students in undergraduate units and ~12% in postgraduate units, with large variation in postgraduate units (none to 34%). Task nature did not appear to drive usage.
• Proactive, high-achieving students used the tool more; the most common submission counts were one, two, or three, with multiple submissions often made in a single day (within ~2 minutes to 5 hours) — a proxy for students acting on and taking up feedback.

Key Findings

• Six key observations from consultations with teachers and tool usage data: proactive high achievers use the tool more; one/two/three submissions are most common; multiple submissions often happen in a single day; students commented on feedback usefulness; students flagged errors/inaccurate feedback; and average feedback ratings were very positive.
• Student-centered framing: The tool removes the judgment value and power structure from the feedback interaction, potentially making students less emotional and more critical and better placed to exercise evaluative judgment of feedback against their own work (cf. Tai et al., 2018).
• Feedback literacy affordances: Use of the tool evidenced dimensions such as acknowledging feedback as a reciprocal process, processing feedback information, and acknowledging/regulating emotions. Notably, engagement (or its absence) appeared driven more by appreciation of feedback for improving writing than by students' capability to process feedback information.
• Not all students acted on these affordances: Because the tool was optional, a relatively small share of students engaged, and teachers must scaffold awareness of feedback literacy dimensions and strategies for enacting them.

Connection to Existing Knowledge Base

• Feedback Loop: The tool creates immediate, actionable feedback loops students can act on and re-submit against (multiple same-day submissions evidence iterative uptake).
• AI Feedback Quality: Students rated feedback positively on average yet also flagged errors/inaccurate feedback, raising questions about quality and trust calibration of automated feedback.
• Formative Assessment: The tool is used for formative, low-stakes writing feedback rather than grading, and the authors position it within enabling learning activities.
• Self-Regulated Learning / Learner Agency: The proposal to make the tool student-facing and degree-wide targets student self-AI Regulation in Education, agency, and evaluative judgment.
• Higher Education: Grounded in a university-wide tertiary pilot across multiple faculties.
• Teaching: Highlights that teachers must scaffold feedback literacy dimensions and strategies — technology does not displace the teacher's design role.
• AI Literacy: Building understanding of what an AI tool can and can't do is central to the proposed student–AI partnership.
• Scaffolding: Drafting-stage templates and teacher guidance are framed as scaffolds for feedback literacy.

Methodological Notes

Strengths include a real-world, multi-faculty, large-scale pilot (~4,000 students) with authentic writing tasks and co-design with an edtech partner. Limitations acknowledged by the authors: the paper is a concise pilot report presenting preliminary insights and reflections rather than full evaluation; further detailed evaluation is required to refine the tool and the learning design surrounding AI for teaching and learning.

What this means for practice

• Instructors. Let students drive the tool themselves: set the micro-level parameters (sentence length, punctuation, grammar, text structure) and have students run it independently, which removes the judgment value and power structure that makes teacher feedback emotionally loaded and lets them exercise evaluative judgment.
• Instructors. Scaffold awareness of feedback literacy dimensions explicitly rather than assuming access is enough — the tool was optional and average usage reached only about 13% of undergraduate and 12% of postgraduate students.
• Designers. Make the tool student-facing and available across all units so each student decides which task and which aspects of writing to seek feedback on, building Learner Agency and self-regulated feedback literacy without adding teacher workload.
• Designers. Add templates for three drafting stages so students encounter feedback's reciprocal nature and the limits of what an AI tool can do, setting up a student–AI partnership they can extend with their teachers.
• Faculty developers. Design the tool around the seven dimensions of the learning-centered feedback literacy framework (Molloy et al., 2020) rather than around workload reduction, since those dimensions are what the pilot can actually evidence.

Limitations

• The paper is a concise pilot report presenting preliminary insights and reflections rather than a full evaluation; the authors state that further detailed evaluation is required to refine the tool and the learning design around it.
• Reach was large (~4,000 students across 29 units at Deakin University, 2021) but engagement was thin: average usage was ~13% in undergraduate units and ~12% in postgraduate units, ranging from none to 34% in postgraduate units.
• There is no comparison condition and no outcome measure of feedback literacy; the evidence is usage data (submission counts and same-day resubmissions) plus consultations with teachers.
• Usage concentrated among proactive, high-achieving students, so the observed patterns may not represent the wider student population.

Citation

Tubino, L., & Adachi, C. (2022). Reconnecting relationships through technology: Developing feedback literacy capabilities through an AI automated feedback tool. In S. Wilson, N. Arthars, D. Wardak, P. Yeoman, E. Kalman, & D.Y.T. Liu (Eds.), Reconnecting relationships through technology: Proceedings of the 39th ASCILITE Conference (e22039). feedback-literacy