In 30 seconds
In a brief interview, Dr. Avriel Epps discusses AI bias, using content recommendations and missing groups in datasets as examples. She says there is no single fix because bias can enter at multiple points in development.
- 1Reinforcing advantageEpps uses gender differences in content recommendations as an example.
- 2No single fixShe says bias can be introduced at multiple points in development.
- 3Missing dataEpps notes that less powerful groups may not be represented in collected data.
Takeaways
- Epps discusses recommendation systems that can reinforce existing advantages.
- She says bias may enter at multiple stages of development.
- Groups with less power may be absent from datasets when their data is not collected.
Read the edited text by chapter
How bias can enter AI
Epps discusses recommendation systems, development decisions and missing data. The short excerpt does not present a complete remedy.
Asked how she defines AI bias, Dr. Avriel Epps discusses a content recommendation system that may prioritize one gender of creator and continue existing advantages.
When Shira Lazar asks how to remove bias, Epps says there is no single solution because bias can be introduced at many points in development. She adds that less powerful groups may be missing from datasets when the people collecting data do not include them.
More from this channel
0:45EN
Shira Lazar· Oct 21, 2026PodcastA Visit to Paris Hilton’s Private SpaA visitor describes a pink spa, sensory treatments and a chilly cryotherapy session.
0:55ENShira Lazar· Oct 19, 2026PodcastTyler Denk on Building a Creator Business Across FormatsGuest Tyler DenkThe Beehiiv founder describes creators combining channels, revenue models, and in-person events.
0:48ENShira Lazar· Oct 12, 2026PodcastA Second Map of the Creator EconomyHost Shira LazarShira Lazar introduces an updated ecosystem map with a section for independent creators.
0:54ENShira Lazar· Oct 3, 2026Podcast · 2 speakersChelsea Clinton on AI Anxiety, Health and GovernanceGuest Chelsea Clinton · Host Shira LazarClinton discusses examples that give her hope and argues for treating AI as a collection of different technologies.
0:38ENShira Lazar· Sep 21, 2026PodcastA Creator’s Visit to an LTK Event in DallasA short event recap highlights programming, brand displays, and conversations with women building creator businesses.
1:01ENShira Lazar· Sep 19, 2026PodcastAffinity 100: A Creator Ranking Focused on Audience ConnectionA co-creator explains why the list emphasizes fandom, trust, and engagement instead of follower counts alone.