Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Shubham Chakraborty has a total of 22 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Cognitively Diverse Multiple-Choice Question Generation: A Hybrid Multi-Agent Framework with Large Language Models

    by Yu Tian, Linh Huynh, Katerina Christhilf, Shubham Chakraborty, Micah Watanabe, Tracy Arner, Danielle McNamara

    Recent advances in large language models (LLMs) have made automated multiple-choice question (MCQ) generation increasingly feasible; however, reliably producing items that satisfy controlled cognitive demands remains a challenge. To address this gap, we introduce ReQUESTA, a hybrid, multi-agent framework for generating cognitively diverse MCQs that systematically target text-based, inferential, and main idea comprehension. ReQUESTA decomposes MCQ authoring into specialized subtasks and coordinates LLM-powered agents with rule-based components to support planning, controlled generation, iterative evaluation, and post-processing. We evaluated the framework in a large-scale reading comprehension study using academic expository texts, comparing ReQUESTA-generated MCQs with those produced by a single-pass GPT-5 zero-shot baseline. Psychometric analyses of learner responses assessed item difficulty and discrimination, while expert raters evaluated question quality across multiple dimensions, including topic relevance and distractor quality. Results showed that ReQUESTA-generated items were consistently more challenging, more discriminative, and more strongly aligned with overall reading comprehension performance. Expert evaluations further indicated stronger alignment with central concepts and superior distractor linguistic consistency and semantic plausibility, particularly for inferential questions. These findings demonstrate that hybrid, agentic orchestration can systematically improve the reliability and controllability of LLM-based generation, highlighting workflow design as a key lever for structured artifact generation beyond single-pass prompting.


  • Open Access

    Cognitively Diverse Multiple-Choice Question Generation: A Hybrid Multi-Agent Framework with Large Language Models

    by Yu Tian, Linh Huynh, Katerina Christhilf, Shubham Chakraborty, Micah Watanabe, Tracy Arner, Danielle McNamara

    Recent advances in large language models (LLMs) have made automated multiple-choice question (MCQ) generation increasingly feasible; however, reliably producing items that satisfy controlled cognitive demands remains a challenge. To address this gap, we introduce ReQUESTA, a hybrid, multi-agent framework for generating cognitively diverse MCQs that systematically target text-based, inferential, and main idea comprehension. ReQUESTA decomposes MCQ authoring into specialized subtasks and coordinates LLM-powered agents with rule-based components to support planning, controlled generation, iterative evaluation, and post-processing. We evaluated the framework in a large-scale reading comprehension study using academic expository texts, comparing ReQUESTA-generated MCQs with those produced by a single-pass GPT-5 zero-shot baseline. Psychometric analyses of learner responses assessed item difficulty and discrimination, while expert raters evaluated question quality across multiple dimensions, including topic relevance and distractor quality. Results showed that ReQUESTA-generated items were consistently more challenging, more discriminative, and more strongly aligned with overall reading comprehension performance. Expert evaluations further indicated stronger alignment with central concepts and superior distractor linguistic consistency and semantic plausibility, particularly for inferential questions. These findings demonstrate that hybrid, agentic orchestration can systematically improve the reliability and controllability of LLM-based generation, highlighting workflow design as a key lever for structured artifact generation beyond single-pass prompting.


  • ReQUESTA: A Hybrid Agentic Framework for Generating Cognitively Diverse Multiple-Choice Questions

    by Yu Tian, Shubham Chakraborty, Linh Huynh, Katerina Christhilf, Micah Watanabe, Danielle McNamara

    This study presents ReQUESTA, a hybrid agentic framework that integrates LLM-powered and rule-based agents to generate multiple choice questions (MCQs) with distinct cognitive focuses: text-based, inferential, and main-idea. To provide an initial validation of the framework, expert raters evaluated 100 ReQUESTA-generated MCQs using a cognitive classification rubric to assess alignment between intended and perceived cognitive categories. The results indicate a high level of agreement between system-assigned and expert-assigned labels (agreement rate = 0.95), suggesting that ReQUESTA can reliably instantiate targeted cognitive distinctions in question generation. These findings offer preliminary evidence of the framework’s capacity to support cognitively diverse and pedagogically meaningful assessment design. ReQUESTA’s modular and hybrid design supports scalable, iterative, and evidence-based assessment development. Future work will include psychometric validation and expansion to additional cognitive categories such as application-level questions.


  • Quality Assessment Through Learning Engineering: An Evaluation Rubric of LLM-Generated Multiple-Choice Questions

    by Katerina Christhilf, Linh Huynh, Yu Tian, Shubham Chakraborty, Micah Watanabe, Danielle McNamara

    Accurately assessing reading comprehension is essential for monitoring student progress, yet creating high-quality multiple-choice questions is time-consuming for instructors. AI-generated questions offer a promising solution but raise concerns about question quality. To address this gap, we developed a rubric for human raters to evaluate the quality of automatically generated comprehension questions. The rubric incorporates established criteria for question design and documented weaknesses of AI-generated items, including awkward, wordy phrasing and conspicuous correct answers. Through iterative testing with two human raters scoring 200 AI-generated multiple-choice questions, we refined scoring categories and scales, achieving weighted kappa reliability of .79 or higher. The final rubric includes seven categories assessing answer correctness, distractor incorrectness, topic relevance, writing clarity, distractor linguistic features, distractor semantic plausibility, and distractor semantic uniqueness. This tool supports future research comparing question-generation systems and exploring automated question-quality evaluation.

  • Automated Run-on Sentence Detection and Correction for Educational Writing

    by Shubham Chakraborty, Yu Tian, Michelle Banawan, Andrew Potter, Linh Huynh, Yoshita Yajjapurapu, Danielle McNamara

    Run-on sentences, including fused sentences, comma splices, and conjunctive adverb misuse, pose a persistent challenge in student writing, undermining both human evaluation and automated analyses in learning environments. Despite their instructional importance, run-ons are underrepresented in major grammatical error correction (GEC) benchmarks. We present a two-stage NLP pipeline for run-on detection and minimal-change correction, designed within the Learning Engineering Framework to improve writing feedback while preserving student voice. Early annotation of 251 student sentences identified 29 potential run-ons, informing our pipeline design and human validation workflows.

  • Transparency in Carbon Credit by Automating Data-Management Using Blockchain

    by Shubham Chakraborty, Abhishek Anand, Divya Kalash, Anupam Srivastava

    Climate change has raised an alarm for world to ponder about and if not stopped can cause serious repercussions. A feasible solution to alleviate greenhouse gas emissions is via a mechanism that is laid out in Kyoto protocol in 1997. It was signed by close to 192 nations thus creating a system to monetize carbon emissions with an aim to build an authority over it. There have been efforts to build a system in previous decade too but there have been certain shortcomings. The major challenges which exist are data manipulation, lack of transparency and substantial costs. So we have come up with the idea of Blockchain which will build a carbon credit ecosystem thus guaranteeing us security, transparency, accessibility to the standardized carbon markets. We will also be using sensors to automate the data to out consortium.


  • Virtual Office Prototype: A Unity-2D work Simulation

    by C P Mallikarjuna Gowda, Anupam Srivastava, Shubham Chakraborty, Anurag Ghosh, Harsh Raj

    In today’s world virtual office have become the talk of the town. In a very less span of time, the number of people using it have grown exponentially. Virtual Office management is a science that requires the use of a scientific tool that can be used for the management of the virtual office. Many tools and applications can be used for the management of the virtual office. The most important of them is a virtual office management application that uses unity and agora.io. The main purpose of a virtual office management application is to increase work effectiveness and to complete standard operating procedures in it. The virtual office management application should be based on a standard operating procedure. The application should have a user interface that is simple and easy to use. Though the problems of missing customer letters, conflicting schedules are there but it can be rectified as we move on. So, to overcome the problem faced by today’s world, we have designed a prototype of virtual office which can be downloaded from Heroku and after creating a server would be ready to use.

  • Development of Information Technology Telecom Chatbot: An Artificial Intelligence and Machine Learning Approach

    by Mallikarjuna Gowda C P, Anupam Srivastava, Shubham Chakraborty, Anurag Ghosh, Harsh Raj

    Nowadays, educators can show case the technology of chatbots in various fields such as teaching and learning. Earlier the resources are almost negligible, in the field of education with the learning design integrated in it. But now with the advancement of technology, chatbot can fill the gap in the teaching landscape too. Using chatbot in the education domain is reassuring, since these bots can point out some of the logistics and diversified issues that a normal class might face. Besides this, with the further advancement of Artificial Intelligence, tech-giants like Apple, Google, and Amazon are providing platforms where building conversation is more focused rather than technicalities of computer programming i.e., to be more specific in natural language processing. In this paper, we have built a rule based chatbot on a platform named Discord, also showcased that how a chatbot can be integrated into other online platforms to counter the challenges faced in teaching. Q/A features are used to get in-depth knowledge about various preinstalled data in chatbot.