COE-XAI - Activities

Name of the Activity: Expert Talk on “XAI for Responsible GenAI Applications”

Date: 4th December 2025

Time: 2:00 pm - 3:00 pm

The expert began the session by explaining the evolution of AI models, moving from traditional machine learning to deep learning and Generative AI. Emphasis was placed on why explainability has become a critical requirement, especially when AI systems are used for high-stakes decision-making.

The session covered:

Overview of Generative AI models such as Large Language Models (LLMs).
Limitations of black-box models and associated risks.
Introduction to XAI techniques such as feature importance, model-agnostic explanations, SHAP, LIME, and attention-based explanations.
Role of XAI in bias detection, error analysis, model debugging, and regulatory compliance.
Discussion on ethical and responsible AI frameworks.
Interactive discussion on research directions and career opportunities in XAI.

Name of the Activity: Expert Talk on “Building Transparent Agentic AI Systems”

Date: 10th September 2025

Time: 2:00 pm - 3:00 pm

The guest talk on “Building Transparent Agentic AI Systems”, which provided deep insights into the design and governance of autonomous, explainable, and trustworthy artificial intelligence. The session began by defining Agentic AI, outlining characteristics such as autonomy, goal orientation, contextual reasoning, and continuous learning. The speaker elaborated on transparency challenges, particularly the “black box” issue , In deep learning–driven agents and its implications for accountability. Participants were introduced to Explainable AI integration through frameworks like LIME, SHAP, and Model Context Protocols that help interpret AI decision-making processes. The discussion also covered ethical and societal implications, emphasizing the need for human-in-the- loop oversight, fairness audits, and transparent documentation practices such as Model Cards and Data Sheets. Several real-world applications were showcased, including transparent agentic AI systems in digital retail, healthcare diagnostics, and smart governance. The session concluded with future directions in the field, envisioning the development of trustworthy, self-auditing AI agents capable of reporting their reasoning chains and confidence scores.

Event Name: Expert Talk on “Understanding XAI, GenAI and Agentic AI Concepts, Tools and Applications”

Date:21/6/2025

Time: 2:00 pm - 3:00 pm

The session began with an introduction to Explainable AI, emphasizing its importance in building trust, transparency, and accountability in AI systems. Key XAI techniques such as SHAP, LIME, and counterfactual explanations were discussed, along with their role in regulated sectors like healthcare, finance, and law. Moving into Generative AI, the speaker highlighted how models like GPT, DALL·E, and diffusion models are transforming content creation, code generation, and design workflows. The discussion covered both the creative potential and ethical challenges of GenAI, including issues like deepfakes, bias, and data provenance.The concept of Agentic AI was then explored, focusing on intelligent systems that act autonomously to achieve user goals. Examples included AI agents capable of task planning, reasoning, and multi-step decision-making. Tools like Auto-GPT, LangChain, and ReAct frameworks were introduced, showing how agentic systems integrate LLMs with memory, tools, and planning strategies.

Event Name: Debate on Ethical Considerations on AI Powered System

Date:26/3/2025

Time: 2:00 pm - 3:00 pm

Objective:

The objective of this debate is to critically analyze the ethical implications of AI-powered systems, addressing concerns such as bias, privacy, accountability, and the impact on human decision-making. The event aims to encourage constructive discussions on the responsibilities of AI developers, policymakers, and users in ensuring ethical AI deployment.

Description:

Students participated in discussion both for and against various ethical concerns in AI systems. Topics of discussion will include algorithmic bias, data privacy, transparency, job displacement, and AI regulation. The format will include structured arguments, rebuttals, and an interactive Q&A session to engage the audience in critical thinking about the moral responsibilities associated with AI technologies.

Outcome:

By the end of the debate, participants and attendees will gain a deeper understanding of the ethical challenges posed by AI-powered systems and potential solutions to mitigate risks. The event will encourage informed decision-making, promote responsible AI practices, and inspire further discussions on ethical AI governance. A key takeaway will be the formulation of ethical guidelines or recommendations for AI development and deployment.

Event Name: Back Chain Coding

Date:05/12/2024

Time: 2:00 pm - 3:00 pm

Objective:

The primary objective of the programming puzzle-solving event was to enhance students' problem-solving skills by challenging them to solve a variety of coding puzzles. This event aimed to foster critical thinking, encourage collaboration, and promote the practical application of programming concepts in a competitive yet engaging environment.

Description:

Participants were presented with a set of programming puzzles, ranging from basic to advanced levels.

Key highlights of the event included:

Variety of Puzzles: Challenges included number-based logic puzzles and string manipulations.

Language Flexibility: Students were allowed to use any programming language of their choice, such as Python, Java, C++, or JavaScript, to solve the problems.

Outcome:

Students were able to code and illustrate theprogramming puzzles

Event Name: Junkyard

Date:11/9/2024

Time: 2:00 pm - 3:00 pm

Objective:

To preprocess unconventional datasets, tools, or resources to solve AI-related challenges.

Description:

Students work on given unconventional datasets or data sources that may not typically be used in AI projects (e.g., scraped data from social media, obscure public datasets) and clean the data. The targeted students are Beginner/Middle level in ML/AI.

Outcome:

Students were able to make use of using python to clean the dataset and were able to illustrate the different methods involved in preprocessing

Event Name:Illuminate AI Poster Presentation on Explainable AI

Date:27/04/2024

Time: 2:00 pm - 4:00 pm

Objective:

To give an opportunity to the students to develop solutions using AI and XAI and provide a platform to

exhibit their work through posters.

Description:

The students are asked to prepare a poster on societal solutions using AI and XAI. Students have prepared posters for the problem statements like Disease Prediction, payment fraud detection etc. Students have explained different machine learning algorithms and explainable AI techniques LIME, SHAP, Anchors used in their work. This gives an exposure to the students to understand how XAI is applied in various domains, including healthcare, finance, and autonomous vehicles.

Outcome:

Students were able to illustrate the Explainable AI techniques by developing the trustworthy models.

Event Name: AI for All - Enhancing User Empowerment with Explainablity

Date:23/3/2024

Time: 10:30 am - 11:30 am

Objective:

To give students an exposure on building trustworthy models using Explainable AI methods.

Description:

The speakers elucidated the concept of XAI, emphasizing its role in making AI systems transparent and understandable to users. She highlighted the need for XAI techniques to enhance trust and facilitate user empowerment.

She has explained different explainableAI techniques LIME, SHAP, Anchors. Case studies and examples were presented to illustrate how XAI is applied in various domains, including healthcare, finance, and autonomous vehicles. Attendees gained insights into how explainable AI techniques are integrated into practical systems to improve transparency and accountability.

Outcome:

Students were able to understand the need of Explainable AI techniques by developing the trustworthy models