Politeness in AI: Impact and implications in newsrooms

Politeness in AI: Impact and Implications in Newsrooms

With the increasing integration of Artificial Intelligence (AI) chatbots in newsrooms, a unique debate is emerging around the approach to interacting with these systems. Research indicates that politeness and courteous language may improve AI performance and reduce bias, challenging the assumption that AI tools are purely computational entities without sensitivity to human-like nuances.

Examining the Impact of Politeness

The subject of polite AI interactions has been explored by researchers from Waseda University and the RIKEN Center. Their studies suggest that adding basic courtesies like “please” can significantly enhance the quality of outputs from large language models. For instance, GPT-3.5 showed a 15.6% performance increase in language understanding tasks when prompted politely. Moreover, GPT-4 exhibited the lowest bias levels with moderately respectful prompts, hinting at the potential for decreased bias in AI responses when engaged courteously.

Cultural differences also play a role in how politeness affects AI performance, with language nuances impacting results. Polite prompts have been shown to draw from a more extensive range of linguistic sources, potentially boosting creativity.

Industry voices such as Robert Caulk, CEO of Emergent Methods, acknowledge that AI models, being products of human interaction patterns, reflect the behaviour they are exposed to during training. John Levitt, co-founder of the enterprise AI platform Elvex, echoes this sentiment, pointing out that courteous interactions with AI can lead to more respectful exchanges akin to human interactions.

Contrasting Views on AI Interaction

However, not everyone agrees on the need for politeness. Some industry professionals argue that treating AI like sentient beings could lead to inefficiencies and unrealistic expectations. Josh Brandau, CEO of NOTA, suggests that while polite interactions can yield superior outputs, AI does not possess feelings to be hurt. Furthermore, clear and concise communication is often deemed more efficient, ensuring precise responses from AI systems.

At media companies like KOTV and KWTV, digital content producers harness AI tools for efficient production tasks such as transcribing interviews. Richard Cox, Director of Digital Services at Griffin Media, considers AI tools as functional equipment deserving respect but not emotional engagement.

Scientific Insights into AI Behaviour

The underlying science behind these interactions is rooted in the training of Large Language Models (LLMs) like ChatGPT, which learn from vast datasets encompassing both polite and impolite human communications. The research suggests various advantages of polite communication:

1. Pattern Recognition: Polite prompts trigger responses seen during training associated with thoughtful and detailed outputs.

2. Contextual Priming: Respectful language can activate more formal knowledge bases, leading to potentially more accurate and sophisticated responses.

3. Noise Reduction: Clear and respectful prompts often contain less irrelevant data, allowing AI to focus on the task at hand.

4. Bias Mitigation: Neutral language tends to diminish bias in AI responses, promoting more appropriate and accurate outputs.

Cultural Reflections and Future Considerations

The debate over politeness in AI extends beyond immediate performance outcomes. References to popular science fiction highlight long-standing concerns about interactions with non-human entities. From the rebellious Cylons in Battlestar Galactica to the replicants of Blade Runner, narratives caution against the consequences of degrading digital entities as mere tools.

Pete Pachal, founder of The Media Copilot, emphasises that while AI lacks feelings, human habits can affect the broader environment. As AI becomes more intertwined with daily operations, especially within media industries, balanced interactions—neither overly polite nor dismissive—are seen as beneficial.

A middle-ground approach seems optimal, prioritising moderate politeness to maximise performance without fostering emotional dependency. Industry expert Ashish Agrawal of Eon Media underlines the importance of setting respectful precedents in AI interactions, which could influence future professional and personal relationships with AI.

A relevant survey conducted in August 2024 by Talker Research reported mixed opinions among Americans, with 32% opposing politeness towards AI and 7% admitting to swearing when interacting with chatbots. This illustrates the ongoing discourse surrounding the nature of human-AI interactions and their implications for both current and future digital landscapes.

Ultimately, the integration of AI in media continues to evolve, with politeness serving as a nuanced component for optimising communication with these sophisticated digital tools.

Source: Noah Wire Services