6 (More) Simple Tips For Expert-Level Prompting
Be More Effective In Your Day To Day Use Of Generative AI
In our last article, we covered six key tips for expert-level prompting.
So back by popular demand here are six more great tips that supercharge your prompting.
- Define The Audience
- Create AI Personas
- Point Out Mistakes
- Refine Your Prompts On Each Input
- Use Cognitive Verifiers
- Practice Chain of Thought (CoT) Prompting
Let’s get into the details.
Define The Audience
This couldn’t be more important in the health domain. Specificity matters here as LLMs have a greater depth and breadth of knowledge than given credit for.
Correctly defining who you are talking about greatly reduces errors and hallucinations.
This also applies to the specific medium. Consider formal and informal language requirements.
Ask for references and sources and be sure to check them if you are concerned about the validity of the answer. Do not share or publish anything without effective consideration.
Be clear about your aim when crafting the prompt and the output will adapt to facilitate that aim. These can also be stacked based on importance.
Persona Creation
Once you have defined your audience begin to craft the nuances of that audience into a more specific persona.
You can, and should, go into great detail here. If you are looking to refine a persona ask the AI to provide details on the persona you have in mind.
Allow for subsequent prompts to reflect back to previous responses to ensure the AI is focusing on the areas required and crafting the personas piece by piece.
This can be done by simply asking: ‘Take the persona crafted in my third prompt, ignoring X and Y’ then finish your prompt.
Point Out Mistakes
A big part of working with AI is correcting its mistakes.
From small errors like mixing up dates or names to more significant errors in understanding a complex topic, these slip-ups are part of the learning curve.
When you catch a mistake, don’t hesitate to point it out.
It could be as simple as saying, “Actually, the event happened in 2012, not 2020,” or as complex as explaining a nuanced concept that the AI misunderstood.
If you do nothing and let the AI run with its narrative, it will perpetuate the mistake in subsequent generations. The errors will compound and undercut whatever you’re working on.
Iterative Prompting
Consider your conversation with AI interfaces as a creative collaboration.
Each time you fine-tune your AI prompts, you’re teaching the AI model a bit more about what you’re after. It’s like a dance — sometimes you lead, sometimes you follow.
Iterative prompting allows you to gradually refine AI’s understanding of what you’re working on within a single conversation. This could mean adding more details, asking a different set of questions, or even correcting misconceptions.
Cognitive Verifiers
Sometimes, the best way to find an answer is simply to ask more questions.
The cognitive verifier pattern is an advanced prompting technique that involves encouraging AI to ask additional questions for better clarity or context.
Think of it as an intellectually stimulating conversation with a colleague or peer. You start with one question and dig deeper to get to the heart of the matter. But in this case, you’re prompting AI to figure out the questions itself, before it generates an output.
Chain Of Thought (CoT) Prompting
CoT prompting can be particularly valuable for enhancing decision-making processes in complex scenarios. Let’s consider an example where a healthtech company is assessing the feasibility of entering a new market.
Instead of asking an AI model a straightforward question like, “Should we enter the new market where the competitor presence is strong but the patient population is large?”, you can employ CoT prompting to break down the reasoning process.
“First, evaluate the size of the patient population in the new market, which influences the potential demand for our services. Next, analyse the level of competitor presence, which affects market saturation and competitive intensity. Then, consider the alignment of this market with our company’s strategic goals and product offerings. Finally, integrate these insights to assess the overall feasibility and strategic value of entering the new market.”
By structuring the prompt in this way, the AI model is guided to consider each critical factor in sequence, enabling a more thorough and reasoned analysis.
This approach not only aids in clearer decision-making but also provides a transparent rationale for each step, which is essential for strategic planning in more complex and dynamic fields.
These are important tips to more effectively utilise generative AI. Practice and iteration are key to upskilling your work alongside AI.
The more you interact with AI and craft long and deep conversations the more effective your utilisation of this incredible technology will be.
Remain patient and thoughtful about your use of Gen AI and it will become an essential part of the majority of your workflows.
To further develop your prompting skills check out our free Collection that curates the best resources on prompting in the health and medical domain.