How Prompting Affects AI Conversation Quality

With the rise of AI chatbots, more people are using them for social interaction. Bored on a Friday night when you were supposed to have plans? In the past, you’d maybe sit and read a book, listen to a podcast, watch a movie, TV series, or even play a video game. Now, people turn to chatbots. They take on these digital intelligences for a good old-fashioned debate. However, many are left disappointed by the response's lack of quality.

The truth is, AI is only as good as the prompt you give it. They say that a crafter is only as good as their tools, right? Well, an AI is only going to be able to provide quality feedback if you offer a quality prompt. Give an AI system a generic or thoughtless prompt, and its responses will be limited. Give an AI system a more detailed, informative prompt, and it can take the idea and run with it.

Learning to use better prompts is critical to making the most of AI conversations. Especially for people using AI for role-playing or debates, prompts need to be handled with great care. How does prompting impact the quality of the conversation? Let’s have a look.

ai prompting

The Clearer The Prompt, The Better The Response

The simplest way to look at AI conversation quality is that it’s all down to that initial prompt. If you arrive with a generic prompt, it’s unlikely to give the AI enough feedback to give you a worthy level of response. Prompts need to be clear, specific, detailed, and include important qualifiers.

Let’s use role-playing as an example. You want to have conversations with an AI about a role-playing experience based around a popular fantasy setting. Let’s use, for this example, The Lord of the Rings. You want to role-play as a soldier of Gondor who has been asked to prepare a battlement in defence of an Orc invasion.

If you give the AI a basic prompt like ‘I’m a Gondor soldier under your command, and you want me to set up defences ahead of an Orc invasion. What do I do? Then you are likely to receive an equally basic response. If you wanted a more informative and engaging response, you would need to feed the AI a prompt that includes more purpose, more names, more features of the setting, etc.

AI tools and chatbots can learn from an incredible array of sources – quickly. Providing the AI chatbot with a more informative prompt gives it more avenues to explore in its research. Thus, the AI bot can come back with a more informative, detailed response based on your level of clarity.

The RTF System

In AI conversation prompt engineering, a common framework is RTF (Role, Task, Format). With an RTF framework, you make it much easier to give the AI:

  • A specific role – the AI now knows who it is supposed to be mimicking in this conversation
  • A clear task – write in specific detail about what you want the AI to carry out in this chat
  • A chosen format – do you want the chatbot to respond in words? Images? What?

RTF is popular for those who are looking to get a chatbot to ‘come out of its shell’, so to speak. You use this to give clear instructions about who the chatbot is supposed to be, what you want it to achieve and what kind of format you would like the bot to respond in. That level of detail ensures prompts are clear and specific, making them more likely to produce satisfying results.

The CRISPE System

Another popular system is CRISPE (Context, Request, Input, Specify, Example).

You set a clear, detailed scenario that the chatbot can use as a basis for its learning. Then, you offer the chatbot a clear request – what you want it to do, basically. You will also need to feed in some key information – returning to our above example, a bit of background about its character for this role-playing event – and provide a detailed, specific rundown of the format, the length of response, etc. – if possible, show the AI chatbot an example of what you are looking for.

This works extremely well when you already have everything you want in place and simply want to use the chatbot as your ‘partner’ for testing out something like a story or screenplay.

Does Saying Please Help AI Conversation Quality?

Yes, it absolutely does! We get it, some people feel a bit odd about using human-like language to talk with a chatbot. However, if we take the idea of politeness beyond just when talking to humans or animals, why not also to an AI?

People often struggle to create high-quality AI chat prompts because they still view it as talking to a computer. So, humanise the conversation. Instead of writing a prompt that looks like a line of code, ask the question the same way you would to a person. Believe it or not, talking ‘nicely’ to an AI – saying please, thank you, etc. – is more likely to result in a more detailed, friendly response.

Prompts that are written nicely, the same way you would ask a colleague or friend, are far more likely to produce the kind of results you are looking for. Prompts are more likely to be more detailed and less stunted creatively if you talk to the AI the way you would a friend or co-worker.

Think about how you would ask someone a question in real life, or request their help. It’s often couched in language that’s full of qualifiers (if you aren’t able to, that’s fine) and understanding if they cannot assist (you don’t need to do X, only Y). Well, that works the same for AI.

Remember, AIs are trained on human interaction. They therefore respond to social niceties and clear, informative instructions better than they would if you ask in a gruff tone and state the bare minimum. Fill in as many gaps as you can with detailed, friendly and respectful prompts and your AI conversation quality is likely to go through the roof!

Home > Blog > How Prompting Affects AI Conversation Quality