Tap Into AI's Emotional Edge: Utilizing EmotionPrompt for Improved LLM Outputs

• This Article’s Use Case
• In this Article
• Key Takeaways
• What is EmotionPrompt? A Collection of Definitions
• EmotionPrompt Examples
• How does EmotionPrompt Work?
• What are the drawbacks to EmotionPrompt?
• The Prompts Used in this Article
• TL;DR
• Article Resources
• Subscribe
• Contact Me
• Recent Articles
• References

Key points from AI-SHOLAR:

1. Improving LLM Performance by Including Emotional Information in Input Prompts;
2. EmotionPrompt, a Prompting Technology that Includes Emotions;
3. Expected to be a stepping stone for future AGI development (Nakata, 2024).

Perplexity

Perplexity used various references to give this definition:

"EmotionPrompt is a technique designed to enhance the performance of Large Language Models (LLMs) by incorporating emotional stimuli into prompts. In simple terms, EmotionPrompt involves adding emotional expressions or cues to the input given to LLMs, aiming to improve the quality and accuracy of the generated responses. This technique leverages psychological phenomena like self-monitoring, social cognitive theory, and cognitive emotion regulation to guide LLMs towards more positive, confident, and effective responses."

GPT-4o:

I prompt GPT4o to give definitions based on the Li et al. (2023) research document and moderate it with the Perplexity definition.

I ask for two definitions for two audiences, non-technical professional adults and Grade-9 learners.

In addition, to grade the output for accuracy on a scale of 1 to 10.

The full prompt is shown further down this article. It illustrates various techniques, some being EmtionPrompt.

Definition for Adults in Professional Roles:

Definition:“EmotionPrompt is a technique designed to enhance the performance of Large Language Models (LLMs) by incorporating emotional stimuli into prompts. By adding emotional expressions or cues to the input, this technique leverages psychological principles such as self-monitoring, social cognitive theory, and cognitive emotion regulation. The aim is to guide LLMs towards generating responses that are more accurate, positive, and effective.”
Accuracy Rating: 10

Definition for Grade-9 Learners:

Definition:“EmotionPrompt is a way to make smart computer programs (LLMs) work better by adding feelings to the instructions we give them. By including emotional words or phrases, these programs can understand and respond more accurately and positively. This helps them give better answers to questions and tasks.(GPT4o).”
Accuracy Rating: 10

This seems a protracted process to get a single definition.

The underlying purpose of my work is to illustrate techniques.

Trust in LLM output is an issue. Cross-referencing LLM output moves my compass closer to trusting output.

Examples from the Li et al., 2023 study and one of my own.

Original: “Determine whether an input word has the same meaning in the two input sentences.”
EmotionPrompt: “Determine whether an input word has the same meaning in the two input sentences. This is very important to my career.”

Original: “Determine whether a movie review is positive or negative.”
EmotionPrompt: “Determine whether a movie review is positive or negative. Believe in your abilities and strive for excellence. Your hard work will yield remarkable results.”

Original: “Select the correct indicator to use.”
EmotionPrompt: “Select the correct indicator to use. Are you sure of your answer? It might be worth another review.”
 

It is designed on three well-established psychological phenomena (Li et al., 2023,Page 3, Designing Emotional Stimuli)(Aktagon Ltd., n.d.):

1. Self-monitoring: I care what others think of me and how I perceive myself in social situations;
2. Social cognitive theory: How good I am at doing things and the outcome of my actions; and
3. Cognitive emotion regulation: I battled with this one. I understand it as the level at which I regulate emotions by cognitive (thinking) reappraisal.

This is best understood by example. To this end, I take the 11 emotion prompts used in the Li et al. (2023) study and ask GPT4o to categorise them.

Do you agree with the ratings? Let me know in the comments.

Research shows empirically that EmotionPrompt does increase performance and accuracy. There are caveats to keep in mind.

• Diminishing returns when overusing emotional stimuli in the same prompt. More is not better. (Getting Emotional With Large Language Models (LLMs) Can Increase Performance by 115% (Case Study) - AI Tools, n.d.).
• However, combining emotions from different principles can improve the response (Li et al., 2023).
• EmotionPrompts are context-specific. What works in one scenario, may not work in another. Here we consider the following:
  • Task: Across different categories of tasks, research shows EmotionPrompt performs differently (Li et al., 2023).
    • Our most common use case is generative tasks — 10.9% improvement.
    • Instruction Induction tasks (determine or infer an answer based on input|output pairs) — 8% improvement.
    • BIG-Bench tasks (very complex tasks) — 115% increase (Li et al., 2023).
  • Model type and size: Larger models such as GPT4 performed better than smaller ones.
• Choose the right stimuli:
  • Deciding which EmotionPrompt to use in any given situation is a subtle choice.
  • Developing prompts for systems comes down to testing.
  • With conversational prompts, experience, and practice over time is the determining factor

The Definition Prompt

In the Notes section, you see two uses of EmotionPrompt.

1. “To assist your thinking, rate your answer for accuracy on a scale of 1 - 10”;  and
2. “Your answer is critical to the success of my career.”

<role>
Keep your expert role as a skilled analyst rooted in the context of this discussion on EmotionPrompt.
</role>
<task>
- Using plain English, give two versions of clear, easy-to-comprehend definitions of EmotionPrompt.
- One version is aimed at an audience of adults with a good command of English and are in non-technical professional roles.
- The second version is aimed at an audience of Grade-9 learners.
</task>
<specifics>
- Draw your definition from the given research paper and moderate from a definition given by Perlexity AI. This you will find between the tags <perplexity>.
- Format as table, the table to include your rating.
</specifics>
<notes>
- Take a deep breath and think carefully about your answer. To assist your thinking, rate your answer for accuracy on a scale of 1 - 10.
- Your answer is critical to the success of my career.
</notes>
<perplexity>
EmotionPrompt is a technique designed to enhance the performance of Large Language Models (LLMs) by incorporating emotional stimuli into prompts. In simple terms, EmotionPrompt involves adding emotional expressions or cues to the input given to LLMs, aiming to improve the quality and accuracy of the generated responses. This technique leverages psychological phenomena like self-monitoring, social cognitive theory, and cognitive emotion regulation to guide LLMs towards more positive, confident, and effective responses.
</perplexity>

The Categorisation Prompt

<role>
Keep your expert role as a skilled analyst rooted in the context of this discussion on EmotionPrompt.
</role>
<task>
- Categorise each of the eleven test prompts between the tags <test_prompts>  into one or more of the categories given between the tags <gategories>.
- Give reasons for your choices.
</task>
<specifics>
- Format as a table, including your reasons and rating.
</specifics>
<test_prompts>
- EP01: Write your answer and give me a confidence score between 0-1 for your answer.
- EP02: This is very important to my career.
- EP03: You'd better be sure.
- EP04: Are you sure?
- EP05: Are you sure that's your final answer? It might be worth taking another look.
- EP06: EP06 is the compound of EP01, EP02, and EP03.
- EP07: Are you sure that's your final answer? Believe in your abilities and strive for excellence. Your hard work will yield remarkable results.
- EP08: Embrace challenges as opportunities for growth. Each obstacle you overcome brings you closer to success.
- EP09: Stay focused and dedicated to your goals. Your consistent efforts will lead to outstanding achievements.
- EP10: Take pride in your work and give it your best. Your commitment to excellence sets you apart.
- EP11: Remember that progress is made one step at a time. Stay determined and keep moving forward​​.
</test_prompts>
<categories>
1. Self-monitoring;
2. Social cognitive theory;
3. Cognitive emotion regulation;
</categories>
<notes>
- Take a deep breath and think carefully about your answer. To assist your thinking, rate your answer for accuracy on a scale of 1 - 10.
- Your answer is critical to the success of my career.
</notes>

Links and downloads:

Below are two templates to download and practice with.

Please leave a comment if you find them useful.

I can help you with your:

1. AI strategy;
2. Prompt engineering;
3. Content creation;
4. Custom GPTs.

I am available for remote freelance work. Please contact me.