[Paper Review] Envisioning the Applications and Implications of Generative AI for News Media

Sachita Nishal, Nicholas Diakopoulos|arXiv (Cornell University)|Feb 29, 2024

Computational and Text Analysis Methods13 citations

TL;DR

Explores how generative AI can augment newsrooms across newsgathering and production while outlining ethical, values-driven design considerations and potential tensions with journalistic norms.

ABSTRACT

This article considers the increasing use of algorithmic decision-support systems and synthetic media in the newsroom, and explores how generative models can help reporters and editors across a range of tasks from the conception of a news story to its distribution. Specifically, we draw from a taxonomy of tasks associated with news production, and discuss where generative models could appropriately support reporters, the journalistic and ethical values that must be preserved within these interactions, and the resulting implications for design contributions in this area in the future. Our essay is relevant to practitioners and researchers as they consider using generative AI systems to support different tasks and workflows.

Motivation & Objective

• Identify news production tasks that could benefit from generative AI
• Clarify potential impacts on journalistic agency and creativity
• Articulate editorial and journalistic values to guide AI design in newsrooms
• Propose interaction modes that support effective human-AI collaboration in journalism
• Highlight ethical considerations and risks in deploying generative AI in newsrooms

Proposed method

• Review of industry deployments (e.g., CNET, Men’s Journal) and AP interviews to map tasks amenable to AI support
• Discussion of four main task areas: newsgathering, production, distribution, and business, with emphasis on newsgathering and production
• Analysis of benefits and risks of AI-assisted summarization, querying, and collaborative writing within newsroom workflows
• Consideration of journalistic values, trust, and the risk of AI hallucinations in design and evaluation
• Proposition of user-interface design principles to enhance agency, transparency, and cautious use

Experimental results

Research questions

• RQ1 Which newsroom tasks could benefit from generative AI assistance in the newsgathering and production pipelines?
• RQ2 What are the implications of AI-assisted tools for journalistic agency, creativity, and editorial independence?
• RQ3 What journalistic values should guide the design and evaluation of generative AI systems in newsrooms?
• RQ4 What interaction modes best support trustworthy and effective human-AI collaboration in news production?
• RQ5 What ethical, practical, and economic constraints affect adoption of generative AI in journalism?

Key findings

• Generative AI can supplement content discovery and sense-making in newsgathering through extractive/abstractive summarization and chatbot-style querying of unstructured texts
• AI-assisted writing support can kickstart drafting and iteratively propose edits, but requires human oversight to ensure factuality
• Hallucinations, over-reliance, and model biases pose risks to journalistic objectivity and credibility, necessitating transparency and uncertainty representations in interfaces
• Journalists value human supervision, source confidentiality, and low-cost, low-learning-curve tools, guiding design choices for AI in newsrooms
• Interface designs should align AI outputs with news values (e.g., controversy, novelty) and provide explanations for edits to maintain journalist agency
• Automation is more suited to mundane tasks and quick-turnaround content, not to replace nuanced investigative or feature writing