File size: 4,144 Bytes
b2c5443
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
87cdc48
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
---
title: Cultural Exposure and Algorithmic Risk Model
emoji: "🧭"
colorFrom: "blue"
colorTo: "green"
sdk: gradio  # THIS IS THE CRITICAL LINE
app_file: app.py
---
# Cultural Exposure & Algorithmic Risk (CEAR) Baseline v1.0

## Model Description

The **Cultural Exposure & Algorithmic Risk (CEAR) Model** is an **analytic, rule-based scoring system** designed to help users and researchers interpret social media usage in terms of its potential impact on cultural awareness and algorithmic vulnerability.

This version is a V1 Baseline: it is **deterministic** (theory-driven by fixed rules and weights) and does not rely on supervised machine learning or proprietary user data.

### 🎯 Key Outputs

1.  **Cultural Connectedness Score (C-Score):** Estimates exposure to viral and trending content, modeled with diminishing returns on time.
2.  **Algorithmic Risk Score (A-Risk):** Quantifies vulnerability incurred from concentrated time on high-intensity, opaque algorithmic feeds.
3.  **Platform Diversity Index (D-Index):** Measures the concentration/spread of usage across platforms (using $1/\text{HHI}$).
4.  **Cultural Efficiency:** Per-platform estimates of C-Score gained per minute spent.

## ⚙️ Analytic Basis & Scoring Logic

The model is defined by transparent assumptions encoded in the Python code (`cear_model.py`) and the platform weights (`platform_weights.json`).

### Core Formulas

The key to the C-Score is the **Diminishing Returns Function** ($f_{DR}$), which prevents the C-Score from increasing linearly with time, acknowledging that the first hour is likely more valuable than the tenth.

$$f_{DR}(\text{Min}) = \log_{10}(\text{Min} + 1)$$

The final scores are calculated as:

$$C_{Score} = \sum_{i} \left[ W_{C,i} \times f_{DR}(\text{Min}_i) \right]$$

$$A_{Risk} = \sum_{i} \left[ W_{A,i} \times \text{Min}_i \right]$$

*(Where $W_{C}$ is the Trend Density Weight and $W_{A}$ is the Algorithmic Risk Weight, defined in `platform_weights.json`.)*

## 🚀 Deployment & Usage (Hugging Face Space)

This repository contains the core logic (`cear_model.py`) and the application interface (`app.py`) for a Hugging Face Space.

### Model Integration (The Engine)

The core logic can be imported and run in any environment:

```python
import pandas as pd
from cear_model import CEARModel

# Example Input Data
user_data = pd.DataFrame([
    {'platform_name': 'TikTok', 'minutes_per_week': 450},
    {'platform_name': 'YouTube', 'minutes_per_week': 200},
    {'platform_name': 'Reddit', 'minutes_per_week': 50},
])

model = CEARModel()
results = model.calculate_scores(user_data)
# {'C_Score': 3.75, 'A_Risk': 565.0, ...}

# Application Interface (The App - app.py)

The app.py script uses the Gradio library to create an interactive web interface. It handles:

    Collecting user input via a table component.

    Calling the CEARModel.calculate_scores() method.

    Generating a qualitative natural language summary based on the quadrant of the C-Score and A-Risk (e.g., "High C, Low A").

⚠️ Limitations and Ethical Considerations

1. Theoretical, Not Validated: The scores are based on fixed, theoretical assumptions about platform design. They are not calibrated against real-world user survey data or outcomes (e.g., actual cultural literacy, actual regret). Scores are relative estimates only.

2. No Content Analysis: The model only uses time and platform. It cannot distinguish between a productive hour watching educational content and an unproductive hour scrolling low-quality content.

3. Future Work: This deterministic model serves as a foundation. Future versions are intended to use the same input schema to train supervised machine learning models that directly predict outcomes (e.g., predicting user-reported "felt caught up" or "post-scroll regret").


---

## 2. `requirements.txt` (For Deployment)

This file lists the necessary Python packages for the Gradio Space to run your model and interface correctly.

```text
# requirements.txt

# Core Model Dependencies
pandas
numpy

# Gradio Space Dependencies
# Gradio is used to build the simple web application interface (app.py)
gradio