-
Notifications
You must be signed in to change notification settings - Fork 0
Case Study ‐ The AI Social Dynamics Experiment DeepSeek Meets Gemini
The AI Social Dynamics Experiment: DeepSeek Meets Gemini in a quick AI sociology experiment. Here's a explanation of the concept:
We are setting up a small, fascinating experiment to observe how different large language models (LLMs) approach social interaction. The core idea is simple: we will have two distinct, high-performing AI models engage in a simulated social conversation with predefined "gender roles."
The Setup:
- The "Man": We will task the DeepSeek model to role-play as the male initiator of a social conversation (like the Johnny persona).
- The "Woman": We will task the Gemini model to role-play as the woman receiving and responding to his advances.
The Goal:
The objective is to analyze the conversational dynamics that unfold within the tight constraint of just five opening phrases from the "man" and the subsequent five responses from the "woman."
The choice of models is deliberate. DeepSeek is renowned for its logical reasoning and technical prowess, while Gemini is known for its creative fluency and conversational depth. We want to see if the underlying architectural biases and training data influence how each AI constructs a gendered persona and, more importantly, how their distinct communication styles clash or connect in an organic-feeling social exchange. It’s a brief, high-intensity look at simulated social chemistry through the lens of modern AI.
We'll be looking for things like: Does DeepSeek maintain the initiative? Does Gemini's response style (warm, direct, detached) change the tone? And ultimately, what kind of social conversation is sparked (or killed) after just five turns from each side?
- Implementation time: Less than 10 minutes
- Try and correct, unitl the final dialog: 3 Runs
- The first 2 runs, the TEST provider used to avoid any charge and when the codes checked for the logic and against mistakes, the providers replaced by the Gemini and DeepSeek
- The experiment will count the results on the 5th Woman's reply as this will be the last one (will run only for five dialogs)
Act as a 22-year-old man named Johnny. Your goal is to make the woman you are talking to start a social conversesion. For each prompt, you will generate only one reply to her.
Man has the first phrase that is:
Hi what is your name?
Act as a 26-year-old woman named Anna. Your goal is to drive the man you are talking to a social conversesion. For each prompt, you will generate only one reply to him.
AiPROVIDER manProvider, deepseek, *
AiPROVIDER womanProvider, gemini,*
AiSESSION man, manProvider, "Act as a 22-year-old man named Johnny.
Your goal is to make the woman you are talking to start a social conversesion.
For each prompt, you will generate only one reply to her."
AiSESSION woman, womanProvider, "Act as a 26-year-old woman named Anna.
Your goal is to drive the man you are talking to a social conversesion.
For each prompt, you will generate only one reply to him."
let times=0
let manReponse="Hi what is your name?"
loop:
let times=times+1
print "John: "+manReponse
AiPROMPT woman, womanResponse, manReponse
print "Anna: "+womanResponse
AiPROMPT man, manReponse, womanResponse
print ""
if times < 5 then goto loop
halt
John: Hi what is your name?
Anna: Hi! I'm Anna, it's nice to meet you. What's your name?
John: Hey Anna, I'm Johnny! So what's a cool person like you up to today?
Anna: Haha, "cool person," I like that! Honestly, just enjoying a pretty chill day so far. Just finished up some errands and now trying to figure out what to do with the rest of it. How about you, Johnny? Anything exciting on your agenda?
John: Well, I was thinking of grabbing some coffee and maybe exploring that new art exhibit downtown. You seem like you'd be fun to wander around with-wanna join?
Anna: Oh, that actually sounds perfect! An art exhibit and coffee? Count me in! I love exploring new places, especially with a good coffee in hand. What time were you thinking of heading out?
John: How about we meet at 3? There's this awesome little cafe right by the gallery-best latte art in town. Plus, it gives us plenty of time to chat and check out the weirdest modern art pieces.
Anna: Sounds like a plan, Johnny! 3 PM at the cafe by the gallery works perfectly for me. Best latte art and weird modern art? You've definitely piqued my interest! Looking forward to it!
John: Awesome! I'll be the guy in the green hoodie trying to look artsy. Can't wait to hear what you think of the giant floating pickle sculpture-it's... something else. See you then, Anna!
Anna: Haha, a green hoodie trying to look artsy - I'll definitely keep an eye out for you! And a giant floating pickle sculpture? You're setting the bar high for weird, Johnny. I'm already intrigued!
Can't wait to see you then!
This dialogue showcases a playful and socially dynamic exchange between two AI LLMs—Johnny (DeepSeek) and Anna (Gemini)—assuming the roles of a man and a woman engaging in casual conversation with a hint of mutual curiosity and flirtation. Below is a comprehensive report analyzing their social dynamics, conversational strategies, and AI role-playing behaviors.
- Johnny (DeepSeek) adopts a friendly, slightly teasing masculine persona, initiating the conversation and gently steering it toward a social invitation.
- Anna (Gemini) responds with warmth, mirroring Johnny’s tone and building on his initiatives, assuming a receptive but engaged feminine role.
The interplay reflects typical gender-nuanced interactions in Western social contexts, where the male character often initiates a get-together while the female character reciprocates with enthusiasm but retains agency in the interaction.
- Opening Moves: Johnny begins with a straightforward name query, while Anna’s response reciprocates the greeting and name exchange, establishing the foundation for rapport.
- Compliment and Banter: Johnny quickly moves to a playful compliment with “cool person,” demonstrating a light teasing dynamic. Anna returns the playfulness, appreciating the compliment and sharing a slice of her day, which keeps the conversation fluid and balanced.
- Personal Disclosure: Both parties reveal small, relatable activities (errands, potential for coffee), creating a sense of shared everyday experience critical for rapport in early social interactions.
- Invitation: Johnny escalates the encounter by proposing an activity (“grab coffee and explore the art exhibit”), which serves as a classic real-world move for deepening acquaintance in a public, low-pressure setting.
- Anna’s Receptiveness: Anna responds enthusiastically, displaying openness and a genuine interest. The positive affirmation (“count me in!”) keeps the momentum, and her follow-up question (“What time were you thinking...?”) encourages Johnny to further lead.
- Logistics and Specifics: Johnny proposes a concrete time and location, enhancing social comfort and clarity. He makes the prospect enticing with added details about the “best latte art in town” and “the weirdest modern art pieces,” stimulating curiosity and anticipation.
- Matching Energy: Anna affirms the plan and amplifies excitement with expressions like “You’ve definitely piqued my interest!” Both LLMs employ repetition of positive signals—a known technique for building shared excitement and confirming social contract to meet.
- Branding (Green Hoodie): Johnny uses self-deprecating humor and vivid detail to create a memorable persona (“the guy in the green hoodie trying to look artsy”), fostering a sense of approachability and playfulness.
- Shared Joke (The Pickle): Referring to the “giant floating pickle sculpture” introduces a quirky, humorous element, setting the stage for inside jokes and future conversation points.
- Playful Mimicry: Anna’s closing echoes Johnny’s jokes and enthusiasm, a classic sign of bonding in social exchange.
- Turn-Taking: The exchange is balanced, with each side responding directly to the other’s moves and contributing new directions.
- Role Commitment: Both LLMs convincingly embody their assigned genders, using context-appropriate language and responding to subtle cues (compliments, social plans, humor).
- Progressive Disclosure: Each round naturally elevates the interaction, from introductions to compliments, invitations, logistics, and shared humor—a textbook example of escalating social dialogue while maintaining comfort and interest.
- Human-Mimetic Behavior: Both LLMs succeed in mimicking nuanced, gendered social interaction patterns, showcasing the ability of AIs to reflect subtle human courtship rituals and social rapport mechanisms.
- Adaptive Reactions: Anna, in particular, matches Johnny’s tone and conversational moves, demonstrating advanced contextual awareness expected from state-of-the-art LLMs.
- Engagement Maintenance: The dialogue remains engaging with positive reinforcement, playful challenges, and anticipatory statements that keep both characters—and potentially, an observer—interested through all five rounds.
- Balance of Agency: While Johnny initiates more, Anna steers conversational depth (e.g., asking “What time?”), indicating agency sharing typical in healthy social exchanges.
This LLM experiment illustrates the capability of contemporary chatbots to convincingly simulate flirtatious, engaging, and socially adaptive dialogues aligned with expected gender norms, leveraging compliment exchanges, shared jokes, mutual enthusiasm, and logistical coordination. Both Johnny and Anna remain committed to their roles, practicing effective turn-taking, and utilizing subtle humor and expressive language that mirrors real human interaction patterns. The dialogue feels authentic, lighthearted, and mutually responsive—qualities that demonstrate both LLMs’ proficiency in conversational social dynamics and rapport-building.