When we talk to people, the tone of our words changes how they respond. The same is true when we interact with Large Language Models. While an LLM doesn’t have feelings, it has been trained on tons of human conversations where politeness, hedging, and formality matter. As a result, the words we choose, such as: “please,” “thank you,” “could you,” or even role instructions—can change how the model replies.
Different types of tone I’ll be exploring:
- Direct tone: short & to the point
- Polite tone: softer phrasing like “could you please…”
- Role assignment: telling the model to act like a “teacher” or “editor” shifts its style to match that role.
Even though the model doesn’t care if we say “please,” it still mirrors polite conversational patterns. This shows how communication habits shape the way language works, even with machines.
The Proof is in the Prompts
Let’s see what happens when we test these prompts with Google: Gemini 2.5 Flash. I know ChatGPT is usually the most popular- so I’m really excited to explore Gemini and other models. The prompts I will be exploring are arbitrarily going to ask about issues with online learning- well I guess maybe I’m a little biased and interested in the subject because I was online for a portion of high school due to Covid-19. That being said you can test this out yourself with any question that interests you as long as it fits the different tone types.
Direct Prompt:
List three problems with online learning
Google: Gemini 2.5 Flash:
Here are three common problems with online learning:
Digital Divide and Technical Issues: Not all students have reliable internet access, suitable devices (like a laptop instead of just a phone), or the technical literacy to navigate online platforms effectively. This creates an equity gap, leading to significant frustration, missed assignments, and an inability to participate fully for those lacking the necessary resources or technical support.
Lack of Engagement, Motivation, and Self-Discipline: Many students struggle to stay focused, motivated, and disciplined when learning independently online. The absence of a structured classroom environment, direct teacher presence, and peer pressure can lead to procrastination, distractions, and a general decline in active participation and self-regulation.
Limited Social Interaction and Collaboration: Online learning often reduces opportunities for spontaneous social interaction with peers and direct, immediate feedback from teachers. This can hinder social-emotional development, collaborative skill-building, and the sense of community that a physical classroom provides, potentially leading to feelings of isolation.
Analysis:
The direct prompt response was clear and well-organized, identifying three major challenges like we asked for. It provided solid explanations for each point, but the tone felt formal and a bit textbook-like. This shows how direct prompts give efficient answers but usually aren’t very personable.
Polite Prompt:
Could you please list three problems with online learning?
Google: Gemini 2.5 Flash:
Here are three problems with online learning:
Difficulty in Assessing and Providing Individualized Feedback: It can be challenging for instructors to accurately gauge a student’s understanding, engagement, and non-verbal cues in an online environment. Providing timely, personalized, and truly effective feedback on assignments and student progress can also be more time-consuming and less nuanced compared to in-person interactions.
Increased Screen Time and Potential for Burnout: Online learning often necessitates prolonged periods in front of a screen, which can lead to digital eye strain, headaches, fatigue, and a general sense of mental burnout for both students and instructors. This can negatively impact overall well-being, concentration, and the effectiveness of learning.
Challenges with Practical, Hands-On, or Experiential Learning: Certain subjects or skills, such as science labs, vocational training, performing arts, or fieldwork, are inherently difficult to replicate effectively in an online format. The lack of physical, hands-on experience can limit the depth of understanding, skill development, and the ability to apply theoretical knowledge in real-world contexts.
Analysis:
Compared to the direct prompt response, the polite prompt answers focus more on student and teacher wellness. They highlight issues like burnout and screen fatigue, which connect to emotional and physical well-being rather than just access or motivation. This shows how a shift in tone can guide the model to consider human-centered impacts of online learning, not only structural or technical challenges.
Role-Based Prompt:
You are an experienced teacher. List three problems with online learning.
Google: Gemini 2.5 Flash:
As an experienced teacher, I’ve seen firsthand how online learning presents unique challenges that often complicate the teaching and learning process. Here are three significant problems:
Difficulty in Monitoring Engagement and Preventing Academic Dishonesty: In a physical classroom, I can read body language, observe students’ work habits, and intervene immediately if I suspect disengagement or cheating. Online, it’s incredibly difficult to ascertain if students are truly focused, if they’re collaborating appropriately (or inappropriately), or if they’re using unauthorized resources during assessments. This makes it harder to ensure academic integrity and accurately gauge genuine understanding.
Challenges in Building Strong Teacher-Student Relationships and Fostering a Cohesive Learning Community: A huge part of effective teaching involves building rapport, trust, and a sense of belonging within the classroom. These relationships motivate students, encourage participation, and allow me to understand their individual needs and struggles. Online, these connections are much harder to forge. Spontaneous interactions are rare, non-verbal cues are often lost, and the sense of shared experience that builds a classroom community is significantly diminished, which can lead to students feeling isolated and less connected to their learning.
Limitations in Providing Real-Time, Nuanced Support and Adapting Instruction: In person, I can instantly spot a confused look, walk over to a student’s desk, offer a quick hint, or re-explain a concept in a different way based on immediate feedback. This ability to differentiate and adapt instruction on the fly is crucial. Online, this real-time, nuanced support is severely hampered. Questions might be delayed, explanations lack the benefit of immediate back-and-forth, and it’s much harder to provide the personalized, just-in-time intervention that can prevent a student from falling behind.
Analysis
The role-based prompt produced a response that centered on the teacher’s perspective, highlighting issues of academic integrity, weakened relationships, and reduced adaptability in online settings. Which are all things I would expect teachers to care about. Unlike the previous answers, this one emphasizes the teacher’s role in maintaining trust, community, and responsive support, making it feel more personal and grounded in lived experience. The language also becomes “I” centered and leads the model to speak from their time as an “experienced teacher”
Conclusion:
Across the three prompts, we can see how tone shapes the focus of the responses: the direct prompt delivered clear structural and technical problems, the polite prompt shifted attention to student wellness and well-being, and the role-based prompt gave insight into the teacher’s challenges and responsibilities. Together, these variations show that communication with LLMs isn’t just about the content of the question, but also about how it’s asked. In practice, this means we can guide LLMs to highlight different dimensions of the same issue—technical, human, or relational—simply by adjusting the framing of the prompt. One last thing I wanted to mention is that I was very surprised with how detailed each of the responses were. I never asked for explanations, just a list- but the model assumed I wanted detailed answers. Could this be due to using a newer model vs an older model? I’m not sure- but maybe thats a question for another blog!
Thanks for reading