Experiment Overview:
Target metrics:
Direct and indirect assessment of basic and advanced cognitive functions, sensory and motor skills, stress resilience, hemispheric integration, cognitive load, attention span, memory, coordination, movement time, reaction time, and task performance. Our specific angle is the availability of AI assistance.
Target demographics:
Educated individuals who are comfortable with technology and familiar with basic AI usage.
Main focus:
We are prototyping intelligent social networks powered by private agents for collaborative learning and development. Unlike centralized "federated learning," our model treats the human and AI as an
evolving single integrated unit. The aim is to maximize autonomous engagement – driven intrinsically rather than by algorithmic manipulation – and achieved by increasing the participants' awareness, specifically their linguistic awareness.
Experiment design:
Participants were given almost no instructions on the experiment flow, only that it tests general intelligence and consists of two main challenges: "About" and "Signup." This is intended to create a little bit of anxiety while providing a "soft anchor."
Access to the site requires the interaction with a moving target, serving as both a basic motor skill test and a filter for automated bots.
The core content is embedded within a narrative ("lore") designed to feel inconvenient or uncomfortable to the target demographic. The environment contains various "distractions" intended to probe curiosity; while some participants may dismiss these as irrelevant, others will attempt to rationalize them (the latter defining our primary group of interest). Neither challenge requires advanced analytical skills; however, they are demanding and require a solid foundation of cognitive and motor skills.
The Entry Point:
The entry point is a cryptic page featuring linguistic puns. While "nerds" will recognize them easily, anyone can consult a private chatbot (access to a local Gemma 4 is provided). The puns are designed to generate conflicting signals and cognitive loops. Certain hints are designed to provoke human responses, while others are tailored for the AI.
Interaction:
Subjects must interact with the page and self-correct when outcomes differ from expectations. Examiners evaluate the ability to extract information, follow sparse instructions, and effectively utilize a chatbot. Identifying at least three "keywords" and at least one correct grants passage to the final step.
Filtering:
This step does not require active engagement to progress; results are only relevant for subjects who complete the "Signup" challenge. Participants are presented with a tailored narrative and examples of chatbot interactions strategically split into two sections. Featuring interactive elements and hidden messages, the narrative is polarizing and emotionally charged. The objective is to assess the participant's ability to discover and process noisy, contrasting information and generate a coherent, concise perspective (POV). The themes used are based on current, real-world "hot topics" to ensure resonance.
Execution:
The instructions are detailed, the execution is the primary challenge. You can cheat your way through, such attempts will be evident in the data.
Outcome:
Participants who successfully navigated this challenge were presented with tools for self-development and self-exploration to play with, and invited to participate in more advanced exercises (interacting with other human+AI units) and become contributors.