(New approach moves beyond accreditation-focused OBE to provide granular visibility
into learning outcomes at student and question level)
A faculty team from ICFAI has developed an AI/ML-based architecture designed to make Outcome-Based Education (OBE) more granular, student-centric and actionable, moving beyond the predominantly accreditation-driven approach to outcome measurement currently used by institutions.
The work, titled “Exploring the Deployment of AI/ML Architecture for Granular Computation of Learning Outcomes in a Foundational Online MBA Course,” was presented by Prof. Prasad R. at the International Conference on Educational Technology and Online Learning (ICETOL 2026) held in Bremen, Germany, from August 17–20, 2026. The conference received more than 500 papers from 54countries and brought together around 900 participants, providing an international platform for research and practice in educational technology and online learning.
The project was co-developed by Rakesh P., Sunitha U. L. and Prasad R., who are also coauthors of the research. From accreditation compliance to improving learning.
The team identifies a fundamental distinction in the way OBE is currently approached. While much of the existing institutional practice and software ecosystem is oriented towards demonstrating outcome attainment for accreditation, the researchers argue that the greater opportunity lies in using outcome data to improve the actual design and experience of learning. “The key question is what the solution is designed for—whether it is primarily to meetaccreditation needs or to improve the quality of learning design and experience. Thelatter includes the former, but is far more powerful because it empowers both the student and the teacher in creating pathways to outcomes,” said Prof. Prasad R., who presented the research at ICETOL 2026.
The AI/ML architecture developed by the team seeks to make that deeper level of visibility possible.The research was conducted across five MBA courses, with 70 students per course, involving 927 questions across six question/assessment types and formative and terminal assessments. The system mapped course outcomes and Bloom's Taxonomy levels at the question level and computed outcomes for individual students, individual questions, question types, evaluation types and the overall course.
Granularity that can reveal where learning is happening—and where it is not Unlike a conventional course-level OBE report, the architecture is designed to provide visibility into the performance of individual students and individual questions. The research found that attainment of the five course outcomes for overall batch per for mance ranged from 56% to 63%, while attainment across the six Bloom'sTaxonomy levels ranged from 43% to 62%. The analysis also identified comparatively lower attainment at the “evaluating” and “creating” levels. Such granular information can potentially help faculty identify gaps between intended outcomes and what students actually achieve, examine the effectiveness of
pedagogyand assessment methods, and redesign courses for subsequent offerings. For students, the approach can provide greater visibility into their performance across different question types and rubric elements, creating opportunities for more target edfeed back and self-directed learning. At the institutional level, it can support quality improvement and OBE reporting while providing deeper insight than accreditation-focused reporting alone.AI as an enabler, not a replacement for faculty. The architecture uses different AI/ML approaches according to the nature of the assessment. The research deployed GPT-based models for multiple-choice questions,a course-configured Custom GPT for long-form descriptive responses and a multi-callagentic pipeline for project evaluation, with the outputs iteratively reviewed and validated by faculty.
The approach is particularly significant because manual computation of granular OBE outcomes is resource-intensive. The team estimates that manually undertaking the process can require approximately 152 hours for a single course, after the initial learning curve, with much of that effort requiring faculty time.
By automating much of this process while retaining faculty involvement in the design
and validation stages, the architecture could make granular outcome analysis feasible
across a larger number of courses.
An institutional approach to AI-enabled OBE: The researchers see the potential application extending beyond individual courses. Their architecture is intended to retain greater institutional and contextual ownership of the data and enable different stakeholders—including students, faculty and academic leadership—to access the level of outcome information relevant to their roles. The broader objective is to create an OBE ecosystem in which outcome measurement becomes part of an ongoing cycle of learning, feedback, course redesign and quality improvement, rather than an activity undertaken primarily to satisfy accreditation requirements.
