[Paper Review] Acceptance of Artificial Intelligence in Journalism

The two papers addressed in this article analyze from different levels how artificial intelligence is accepted and interpreted in the journalism domain. While the 2023 paper sought a meeting point between press trust and AI trustworthiness at the theoretical level, the 2025 paper empirically reveals the structural conditions for AI utilization through journalists' individual perceptions and acceptance attitudes. In an environment where generative AI is flooding and so-called 'AI slop' has become normalized, these two studies prompt us to again ask how journalism can recover and reconstruct transparency and trustworthiness. Factors Affecting Journalists' Acceptance of AI Technology insightfully shows that journalists' AI use lies not as an individual attitude problem but in the structural tension of organizations and institutions. On the other hand, Trustworthy AI-Based Journalism AI attempts to theoretically formalize commensurability through the conceptual frameworks of AIX, TAI, and JAI, but leaves somewhat to be desired in recommendations for actually implementing commensurability.

Current Status of AI Adoption in Journalism and Awareness Gap Between Organization and Individual

Globally, a clear trend exists where major media organizations like the Financial Times, AP, and Guardian are establishing generative AI guidelines and actively introducing technology into newsrooms. These changes are occurring across all domains of journalism from data collection to article writing and fact-checking, and domestic media organizations are also participating through AI anchors and curation services. However, behind this technological diffusion, expectations about opportunities AI will bring coexist with concerns and resistance about crises such as job replacement, errors, and bias. Particularly noteworthy is that a considerable gap exists between media organization-level active investment and actual constituent journalists' acceptance attitudes. According to the Korea Journalists Association's survey, the utilization rate of generative AI among active journalists is only around 20%, suggesting that technology adoption is closely linked not just to building physical environments but also to individual journalists' perceptions, capabilities, and organizational support systems.

Research's Theoretical Framework: TAM Extended to Reflect Journalism's Specificity

This research was based on Davis's Technology Acceptance Model (TAM) but extended it to reflect the specificity of journalists as a professional group. In addition to the core variables of existing TAM — 'perceived usefulness' and 'perceived ease of use' — 'AI quality perception,' 'journalism quality improvement expectation,' 'AI literacy,' and 'AI efficacy' were set as antecedent variables to analyze structural relationships. This considered that journalists, unlike general technology users, are a group pursuing the professional values of public good and truthfulness. Researchers surveyed 300 domestic journalists and verified causal relationships between variables through Structural Equation Modeling (SEM). In particular, borrowing information quality research and self-efficacy theory, the research aimed to clarify through which paths the reliability of technology perceived by journalists and their ability to utilize technology affect actual acceptance intention.

'Journalism Quality Improvement' Is the Core Driver, More Than Convenience

Technical requirements (4 core requirements): To technically implement AI ethics standards, **diversity respect (bias elimination), accountability (guaranteeing responsibility for results), safety (risk management and security), and transparency (explainability and traceability)** must be satisfied.

However, there are practical difficulties in applying these guidelines to actual journalism settings. AI developer labor costs are high making it difficult for media organizations to bear, and TAI guidelines specialized for the journalism domain are still absent. Also, the problem of 'incommensurability' where AI ethics guidelines are abstract and developers have difficulty implementing them in code still exists.

Human-in-the-Loop and the Role of Journalists

Technical guidelines alone are insufficient, and the research emphasizes the 'Human-in-the-loop' approach. This means humans intervening in the decision-making process of AI systems.
Particularly in JAI, journalists should intervene throughout the AI lifecycle as planners and reviewers, not simply data labelers. This connects with the perspective of 'human-centered AI' and suggests that journalists must become subjects who critically verify AI outputs and project journalistic values rather than blindly accepting AI's results. Furthermore, expansion to 'Society-in-the-loop' with participation of civil society and regulatory authorities is also proposed.

Proposal for TAI-Based JAI

The researcher finally proposes a 'TAI-based JAI' model based on commensurability between press trust and AI trustworthiness. This model must satisfy the following three conditions.

AIX Condition: Practical integration with AI in the journalism field must occur comprehensively.
TAI Condition: JAI must secure technical trustworthiness (transparency, safety, diversity, accountability) from the TAI perspective, and journalists must participate in the process.
Domain Condition: This technical trustworthiness must necessarily contribute to improving journalism's trustworthiness and audience trust.