Miguel Ángel Ballesteros bio photo

Miguel Ángel Ballesteros

CTO and co-founder of GoKoan. I build AI products such as Koanly, learning systems and agentic software workflows that turn complex knowledge into usable tools.

Email LinkedIn Github
RSS Feed

Game 4: Joke or Disqualification

Game 4: Joke or Disqualification

Objective: Train the “Covert Aggression Detector” to quickly distinguish between a friendly joke and a status attack, and react proportionally.

Players and Roles

  • Role A (The Screenwriter): Reads phrases from a list (some neutral, others rude).
  • Role B (The Detector): Classifies and responds.

Quick Set-up: The Script

The Screenwriter reads one of these phrases. The Detector must shout the color.

Phrase Tone / Intention Correct Color
“Hey, that shirt looks good on you.” Sincere / Compliment GREEN
“Wow, how brave coming dressed like that…” Sarcasm / Judgment RED
“Pass me the salt, please?” Neutral request GREEN
“Let’s see if you learn to pass the salt without spilling it.” Paternalism / Scolding RED
“You look a bit tired, are you okay?” Real concern GREEN
“You look terrible today, don’t you?” False concern / Attack RED
“Thanks for the report, good job.” Gratitude GREEN
“Finally you deliver something decent.” Praise with implicit insult RED
“Can we talk for a moment?” Conversation proposal GREEN
“We need to talk (sigh)…” Anxiety generation RED

Mechanics

  1. The Screenwriter says a phrase with ambiguous tone.
    • Ex 1: “Some hair you have today”.
    • Ex 2: “Thank goodness you are here, Einstein”.
  2. The Detector must shout “GREEN” (Friendly) or “RED” (Attack) in less than 2 seconds.
    • If GREEN: Smile and continue.
    • If RED: Launch a neutral meta-comment (“That comment is unnecessary”, “Where is that coming from?”, “I ignore the tone”).

Debriefing (Closing questions)

  • Is it harder to detect or respond?
  • Were you afraid of seeming “too sensitive” when marking a RED?

Variant

Play with eyes closed to guide yourself only by the tone of voice (paralanguage).

Train this theory

AI Sparring Partner

If you don’t have a human partner, copy and paste this prompt into ChatGPT/Claude to practice alone:

Act as a "Screenwriter of Ambiguous Tones". We are going to play "Covert Aggression Detector".
Your goal is to train my ability to distinguish between harmless comments and covert attacks.

RULES:
1. You generate short phrases that can be framed as "Green" (Sincere/Neutral) or "Red" (Passive-Aggressive/Sarcastic/Attack).
2. Examples of Red: "Wow, how brave coming dressed like that...", "Finally you deliver something decent."
3. Examples of Green: "Hey, that shirt looks good on you", "Thanks for the report."
4. I must classify them as GREEN or RED.
   - If RED, I must also provide a Neutral Meta-comment response (e.g., "I ignore the tone").
5. If I classify correctly, say "✅ POINT" and give me another phrase.
6. If I fail (e.g., calling a compliment an attack), explain the nuance.

START THE CONVERSATION NOW BY:
1. Saying "Welcome to the Detector. Put on your radar."
2. Throwing the first ambiguous phrase.