Alfred Ongere Alfred Ongere @ Impact_chaser Telecoms engineer - PowerPoint PPT Presentation

Alfred Ongere

Alfred Ongere @ Impact_chaser Telecoms engineer Intel Software Innovator Team lead & Founder – Ai Kenya

About AiKenya • Ai Kenya is a community of practitioners and enthusiasts in machine learning & Data science across East Africa that come together to share knowledge & challenges in the field. • This is done through meetups, events, online discussion forums and a podcast show.

Algorithmic bias Occurs when a computer system reflects the implicit values of its human designers in its behavior

Illustrative examples of algorithmic bias

May 2009 Facial recognition software by Nikon wrongly assumes people of Asian descent are blinking

June 2015 Google photos tagging people as animals

May 2016 - Biased recidivism against blacks COMP AS (Correctional Offender Management Profiling for Alternative Sanctions )

Bias may be a human problem, but amplification of bias is a technical problem — a mathematically explainable and controllable by product of the way models are trained.

Areas to pay attention to, to minimize bias and observe ethics • Data collection – Minimize sampling bias • Data handling (GDPR) • T erms and conditions – permissions (how data collected is used)

How can we reduce bias in products and services we offer? 1. Advocate for explainability in models (xAI) 2. Involvement in Open audits (sharing datasets and requesting for feedback) 3. Investing in R&D 4. Open source is your friend

T ools and resources to tackle bias AI Explainability 360 T oolkit

Recommended readings

Contact : alfred@kenya.ai Founder & T eam lead, Ai Kenya LinkedIn : Alfred Ongere