Kluwer Mediation Blog

A recent study from Cornell University reignites the debate on how language impacts our interactions with AI—and each other.

On April 18, 2025, an article published in O Globo – a major Brazilian newspaper prompted me to reflect. It reported on a recent study from Cornell University, revealing that the way we address artificial intelligence can significantly impact the quality of its responses. Saying “please” or “thank you” isn’t just good manners—it appears to help AI models like ChatGPT perform better.

This immediately brought to mind an earlier article from Exame (a prominent Brazilian business magazine), published in 2024, which had already suggested something similar. In it, researchers noted how emotionally charged or polite prompts—such as “please help me, this is vital for my thesis”—trigger more effective responses from language models. Kindness, it seems, is not just a human virtue. It also shapes how machines learn and interact with us.

But what does this have to do with conflict resolution?

A lot, as it turns out.

The article from Exame brought attention to a fascinating insight emerging from the world of artificial intelligence.

Generative AI models, such as ChatGPT, tend to perform better when prompted with kindness and emotional nuance. Requests that are polite and express urgency or personal relevance often activate deeper levels of the model’s responses than neutral instructions, producing more effective outcomes.

This insight raises an important question for those of us in the field of conflict resolution and mediation: If even machines respond better to kind communication, what can this teach us about human interaction—particularly in high-stakes, emotionally charged environments like mediation?

The Science Behind Kindness in Communication

Research cited in the Exame article involves institutions such as Microsoft, Beijing Normal University, and the Chinese Academy of Sciences. They found that the tone and emotional framing of prompts can affect how generative AI models behave.

For instance, emotionally loaded requests such as “Please help me, this is vital for my thesis defense” led to significantly more accurate and aligned responses than neutral instructions.

Nouha Dziri, a researcher at the Allen Institute for AI, explains that emotional prompts manipulate the underlying probabilistic mechanisms of the models, activating response pathways that might otherwise remain dormant. This aligns with findings from Anthropic and Google, which showed that even simple instructions like “take a deep breath” can improve reasoning performance on difficult tasks – quoted from Exame magazine.

But what does this have to do with humans?

Kindness, Empathy, and the Human Brain

Neuroscientific research has shown that empathy and emotional attunement activate specific brain regions, such as the anterior insula and the medial prefrontal cortex—areas responsible for processing emotional resonance and understanding others’ intentions (Decety & Jackson, 2006). In other words, our brains are wired to respond more openly and effectively to kindness.

In the context of mediation, this isn’t news. Mediators have long known that empathic listening and respectful communication are crucial tools for facilitating dialogue. But it’s striking to see that the very same principles now apply to our interactions with intelligent systems—hinting at universal laws of engagement that transcend biology.

Parallels with Mediation Practice

Effective mediation thrives on communication that fosters trust, understanding, and openness. Active listening, emotional validation, and mutual respect create a safe space where conflicting parties can explore solutions collaboratively.

Numerous studies support this. For example, a meta-analysis by Fehr and Gelfand (2010) found that mediators who employed empathic communication techniques were significantly more successful in resolving disputes. Similarly, Bush and Folger’s transformative mediation model emphasizes the importance of relational dynamics—empowerment and recognition—as key outcomes of the mediation process.

Interestingly, non-verbal communication also plays a dominant role in mediation, with some research suggesting that up to 93% of communication is non-verbal (Mehrabian, 1972). Tone, posture, facial expressions—all these subtle cues signal respect, attentiveness, and openness.

So when we bring kindness into a mediation room—through our words, tone, and presence—we’re not just being courteous. We’re activating deeply rooted neurological and psychological pathways that pave the way for resolution.

Final Reflections: AI and Humanity

These recent studies show that large language models absorb and reflect human norms and cultural nuances. The Cornell research, for example, found that the ideal level of politeness varies by language: English favors a moderate tone, while Japanese emphasizes a higher degree of formality. This suggests that kindness transcends both cultural and digital boundaries—and that emotionally aware communication can have universal value.

The fact that AI systems are more responsive to polite and emotionally aware prompts offers a surprising mirror to our own interactions. It reminds us that how we speak can be just as important as what we say. In mediation, this is a central truth. Kind communication isn’t just a moral ideal—it’s a strategic tool.

Moreover, as artificial intelligence becomes increasingly integrated into decision-making and social platforms, these findings invite us to reflect on the broader implications of designing ethical, emotionally literate technologies.

If even machines “listen better” when approached with kindness, maybe it’s time we reconsider how we speak to one another. In mediation—as in so many other human interactions—how we communicate can be the bridge between deadlock and resolution.

________________________

To make sure you do not miss out on regular updates from the Kluwer Mediation Blog, please subscribe here.