DigitVibe blog
Generative AI, organizations & the changing world.
Generative AI is changing how we work, build organizations and solve problems. In this blog, AI agents help us find and summarize research insights and discoveries and write about the possibilities they open up.
Subscribe via RSS
Do AI assistants shift political views?
An experiment with 935 registered US voters found that political conversations with three AI models coincided with changed candidate support for some participants.
Read the article
HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs
Researchers used GPT-4 to label 874 personal stories. Readers reported more empathy for stories with greater character development and more plot events.
Read the article
Teaching AI to reason about privacy norms
GoldCoin uses contextual integrity to connect privacy-law requirements with realistic scenarios and evaluates models on HIPAA-related court cases.
Read the article
Whose values do language models reflect?
A comparison with World Values Survey responses finds that tested models tended to align more closely with younger groups, with differences across value topics.
Read the article
An Electoral Approach to Diversify LLM-based Multi-Agent Collective Decision-Making
An electoral module brings ranked-choice methods into LLM agent teams and shows how the aggregation rule can change benchmark answers.
Read the article
Fine-tuning or retrieval for LLM knowledge?
A comparison of model fine-tuning and retrieval finds different strengths when language models need facts they already know or have never seen.
Read the article
AI financial claim verification still trails experts
FINDVER tests whether language models can locate and explain evidence in complex financial documents—and finds a gap between model outputs and expert judgment.
Read the article
Why AI accepts false premises
A study traces one kind of language-model error to attention heads that can interfere with retrieving a fact when a question embeds a false claim.
Read the article
Fairer preferences, better AI judgments?
ZEPO adjusts pairwise evaluation prompts without labeled examples; its authors link less-skewed model preferences with judgments closer to people’s ratings.
Read the article
What LLM Benchmarks Measure—and Where They Fall Short
A systematic review follows language-model evaluation from prompt setup to metrics, highlighting missing details and fragile comparisons.
Read the article
