AI IRELANDPodcast hub
Back to all episodes
AI Ireland Podcast | E176 | Responsible AI and governance

'Ethical AI and Bias Mitigation' with University of Limerick's Deepak John Reji

Deepak John Reji | University of Limerick (association stated; employment not established)

The big idea

Bias mitigation requires representative data, diverse human judgement and continuous evaluation within the specific social context of use.

About this conversation

Today’s guest is Deepak John Reji, Senior Data Scientist, and associated with University of Limerick. Discussing Deepak’s AI Awards 2024 nomination in the Best Use of Responsible AI an Ethics category, this episode centres around the challenges and opportunities in creating safe and responsible large language models (LLMs) that strive to reduce bias and bolster ethical AI practices.

Editorial discussion prompts

Define the harm and affected groups.
Use diverse annotators and record disagreement.
Test cultural and domain variation.
Evaluate against clear baselines.
Monitor fairness after deployment.

Prompts to explore

Define the harm and affected groups.
Use diverse annotators and record disagreement.
Test cultural and domain variation.

Chapter guide

0:00Mitigating bias in AI through fine-tuned LLMs
5:23Challenges in curating unbiased data and training models
8:39How bias perception varies by context, demographics and culture
11:29How diverse annotators ensured culturally representative bias assessments
12:53Enabling scalable, domain-specific LLM solutions through Instruction-led fine-tuning
15:21Focus on responsible AI, integrating ethics and cognitive science

Editorial draft based on the official RSS description. The accompanying machine transcript is available for wording verification; full recording review remains pending. Names, quantitative claims, time-sensitive statements, rights and uncertain transcript wording remain publication holds.

The machine transcript and downloadable workbook remain in editorial review. They will be added only after names, terminology, rights and file delivery are checked.