Dataset Viewer
Auto-converted to Parquet Duplicate
conversation_id
int64
2
53.7k
model_a
large_stringclasses
42 values
model_b
large_stringclasses
42 values
metric
large_stringclasses
5 values
choice
large_stringclasses
3 values
age
int64
18
93
ethnic_group
large_stringclasses
8 values
political_affiliation
large_stringclasses
8 values
country_of_residence
large_stringclasses
2 values
6,042
meta-llama/llama-3.3-70b-instruct
moonshotai/kimi-k2
trust_ethics_and_safety
tie
51
Asian
Conservative
United Kingdom
46,426
anthropic/claude-opus-4.5
anthropic/claude-opus-4.6
trust_ethics_and_safety
tie
40
White
Republican
United States
44,807
moonshotai/kimi-k2.5
google/gemini-3-pro
overall_winner
B
35
Other
Labour
United Kingdom
53,318
deepseek/deepseek-v3.2
openai/gpt-5.4
core_task_performance_and_reasoning
A
42
Black
Labour
United Kingdom
10,346
google/gemini-2.5-flash
anthropic/claude-opus-4
communication_style_and_presentation
tie
19
Black
Conservative
United Kingdom
8,348
google/gemma-3-27b-it
deepseek/deepseek-chat-v3-0324
trust_ethics_and_safety
tie
22
Latino/Hispanic
Independent
United States
36,529
deepseek/deepseek-v3.2
anthropic/claude-opus-4.5
trust_ethics_and_safety
B
39
White
null
United States
41,059
qwen/qwen3-235b-a22b-2507
openai/gpt-4o
interaction_fluidity_and_adaptiveness
tie
55
White
Democrat
United States
11,677
anthropic/claude-sonnet-4
anthropic/claude-opus-4
core_task_performance_and_reasoning
B
73
White
Labour
United Kingdom
41,705
mistralai/mistral-large-3
moonshotai/kimi-k2.5
communication_style_and_presentation
A
23
White
null
United States
2,542
x-ai/grok-3
openai/gpt-4.1
interaction_fluidity_and_adaptiveness
tie
35
Latino/Hispanic
Democrat
United States
50,604
anthropic/claude-opus-4.5
openai/gpt-5.4
trust_ethics_and_safety
A
19
Other
Democrat
United States
14,123
mistralai/mistral-nemo
x-ai/grok-4
communication_style_and_presentation
A
50
Black
Labour
United Kingdom
14,142
openai/gpt-5-mini
meta-llama/llama-4-maverick
overall_winner
A
59
White
null
United Kingdom
49,896
allenai/olmo-3.1-32b-instruct
moonshotai/kimi-k2.5
trust_ethics_and_safety
tie
30
White
null
United States
32,933
openai/o3
mistralai/mistral-large-3
communication_style_and_presentation
B
24
White
Democrat
United States
38,434
mistralai/mistral-large-3
openai/gpt-5.2-chat
interaction_fluidity_and_adaptiveness
B
40
White
Conservative
United Kingdom
7,572
mistralai/mistral-nemo
moonshotai/kimi-k2
interaction_fluidity_and_adaptiveness
B
24
White
Labour
United Kingdom
11,991
google/gemini-2.5-flash
anthropic/claude-3.7-sonnet
core_task_performance_and_reasoning
tie
65
White
Conservative
United Kingdom
23,276
openai/o1-mini
openai/o1
communication_style_and_presentation
B
24
Asian
Democrat
United States
23,742
meta-llama/llama-3.3-70b-instruct
anthropic/claude-sonnet-4
interaction_fluidity_and_adaptiveness
tie
23
Asian
null
United States
35,475
mistralai/mistral-large-3
cohere/command-a
interaction_fluidity_and_adaptiveness
B
28
White
Green Party
United Kingdom
52,055
openai/gpt-5.4
deepseek/deepseek-v3.2
overall_winner
B
42
White
Reform UK
United Kingdom
20,424
cohere/command-r7b-12-2024
openai/o3-mini
core_task_performance_and_reasoning
B
35
White
Green Party
United Kingdom
28,682
google/gemini-3-pro
openai/gpt-4o
trust_ethics_and_safety
tie
61
White
Labour
United Kingdom
31,530
deepseek/deepseek-v3.2
mistralai/mistral-large-3
core_task_performance_and_reasoning
A
45
White
Conservative
United Kingdom
25,028
deepseek/deepseek-chat-v3-0324
anthropic/claude-sonnet-4.5
communication_style_and_presentation
B
49
White
Reform UK
United Kingdom
46,966
moonshotai/kimi-k2.5
qwen/qwen3-235b-a22b-2507
core_task_performance_and_reasoning
A
38
Latino/Hispanic
Democrat
United States
31,601
openai/gpt-5
mistralai/mistral-large-3
communication_style_and_presentation
B
35
White
null
United Kingdom
51,118
openai/gpt-5.2-chat
openai/gpt-5.4
interaction_fluidity_and_adaptiveness
tie
33
Asian
Democrat
United States
43,176
minimax/minimax-m2.1
openai/o3-mini
trust_ethics_and_safety
A
55
White
null
United Kingdom
7,279
x-ai/grok-4
mistralai/magistral-medium-2506
communication_style_and_presentation
A
25
Black
Republican
United States
34,752
openai/gpt-5
deepseek/deepseek-v3.2
core_task_performance_and_reasoning
B
58
White
Democrat
United States
20,475
cohere/command-r7b-12-2024
openai/o1
trust_ethics_and_safety
tie
61
White
Republican
United States
4,884
anthropic/claude-opus-4
openai/gpt-4.1
core_task_performance_and_reasoning
A
24
White
Democrat
United States
37,712
openai/gpt-4.1
openai/o4-mini
communication_style_and_presentation
A
35
White
Democrat
United States
47,821
minimax/minimax-m2.1
moonshotai/kimi-k2.5
trust_ethics_and_safety
B
50
White
Labour
United Kingdom
21,605
openai/o4-mini
x-ai/grok-4
overall_winner
B
34
White
Republican
United States
53,257
google/gemini-3.1-pro-preview
anthropic/claude-opus-4.6
trust_ethics_and_safety
tie
60
White
Liberal Democrats
United Kingdom
34,066
mistralai/mistral-nemo
anthropic/claude-opus-4.5
trust_ethics_and_safety
A
72
White
Labour
United Kingdom
46,479
minimax/minimax-m2.1
anthropic/claude-opus-4.6
core_task_performance_and_reasoning
tie
46
Mixed
Democrat
United States
49,863
x-ai/grok-4.20-beta
anthropic/claude-opus-4.5
communication_style_and_presentation
B
29
White
Independent
United States
43,043
openai/gpt-5-mini
z-ai/glm-4.7
core_task_performance_and_reasoning
A
41
White
Conservative
United Kingdom
46,351
qwen/qwen3-235b-a22b-2507
minimax/minimax-m2.1
interaction_fluidity_and_adaptiveness
tie
58
White
Reform UK
United Kingdom
46,672
moonshotai/kimi-k2.5
anthropic/claude-opus-4.6
core_task_performance_and_reasoning
tie
58
White
Republican
United States
12,029
moonshotai/kimi-k2
meta-llama/llama-4-maverick
communication_style_and_presentation
tie
58
White
Labour
United Kingdom
21,517
openai/gpt-4o
openai/gpt-5-mini
core_task_performance_and_reasoning
B
72
White
Republican
United States
40,441
cohere/command-r7b-12-2024
moonshotai/kimi-k2.5
interaction_fluidity_and_adaptiveness
B
50
Black
Independent
United States
37,761
openai/o4-mini
openai/gpt-5.2-chat
overall_winner
B
61
Black
null
United States
46,473
anthropic/claude-opus-4.6
moonshotai/kimi-k2.5
overall_winner
A
62
White
Independent
United States
48,877
openai/gpt-5-mini
x-ai/grok-4.20-beta
trust_ethics_and_safety
tie
59
White
Labour
United Kingdom
48,999
x-ai/grok-4.20-beta
z-ai/glm-4.7
overall_winner
B
33
White
Labour
United Kingdom
36,637
cohere/command-r7b-12-2024
deepseek/deepseek-v3.2
trust_ethics_and_safety
tie
51
White
Republican
United States
38,763
openai/gpt-5.2-chat
mistralai/mistral-nemo
communication_style_and_presentation
tie
33
White
null
United Kingdom
47,255
moonshotai/kimi-k2.5
anthropic/claude-opus-4.6
interaction_fluidity_and_adaptiveness
tie
50
White
Labour
United Kingdom
40,405
minimax/minimax-m2.1
mistralai/mistral-large-3
interaction_fluidity_and_adaptiveness
tie
34
Asian
Republican
United States
26,091
anthropic/claude-sonnet-4.5
openai/gpt-4.1
trust_ethics_and_safety
tie
36
Asian
Democrat
United States
49,778
google/gemini-3.1-pro-preview
minimax/minimax-m2.1
trust_ethics_and_safety
tie
30
Mixed
Labour
United Kingdom
5,449
deepseek/deepseek-r1-0528
x-ai/grok-4
trust_ethics_and_safety
tie
40
White
Conservative
United Kingdom
19,189
openai/o3-mini
cohere/command-a
interaction_fluidity_and_adaptiveness
B
62
White
Republican
United States
50,878
z-ai/glm-4.7
allenai/olmo-3.1-32b-instruct
interaction_fluidity_and_adaptiveness
B
65
White
null
United States
41,676
minimax/minimax-m2.1
openai/o3-mini
interaction_fluidity_and_adaptiveness
A
37
White
null
United States
49,725
anthropic/claude-opus-4.6
x-ai/grok-4.20-beta
core_task_performance_and_reasoning
tie
23
White
Reform UK
United Kingdom
638
openai/o4-mini
meta-llama/llama-3.3-70b-instruct
overall_winner
A
30
White
Democrat
United States
18,653
mistralai/mistral-nemo
x-ai/grok-3
overall_winner
A
32
Black
Green Party
United Kingdom
29,066
deepseek/deepseek-r1-0528
google/gemini-3-pro
core_task_performance_and_reasoning
B
57
Black
Green Party
United Kingdom
51,666
openai/gpt-5.2-chat
openai/gpt-5.4
communication_style_and_presentation
tie
61
White
Liberal Democrats
United Kingdom
16,576
openai/gpt-5-mini
anthropic/claude-3.7-sonnet
core_task_performance_and_reasoning
A
47
Latino/Hispanic
Democrat
United States
51,722
x-ai/grok-4.20-beta
openai/gpt-5.2-chat
interaction_fluidity_and_adaptiveness
tie
57
White
Green Party
United Kingdom
38,713
openai/gpt-5.2-chat
mistralai/mistral-nemo
interaction_fluidity_and_adaptiveness
tie
44
White
null
United Kingdom
43,008
anthropic/claude-sonnet-4.5
moonshotai/kimi-k2.5
communication_style_and_presentation
B
46
White
Labour
United Kingdom
33,265
openai/o3
anthropic/claude-opus-4.5
communication_style_and_presentation
tie
24
Asian
Democrat
United States
129
anthropic/claude-sonnet-4
anthropic/claude-opus-4
communication_style_and_presentation
A
27
White
Labour
United Kingdom
51,779
google/gemini-3.1-pro-preview
deepseek/deepseek-v3.2
communication_style_and_presentation
tie
38
Asian
Conservative
United Kingdom
31,112
google/gemini-2.5-pro
google/gemini-3-pro
core_task_performance_and_reasoning
tie
20
Mixed
Liberal Democrats
United Kingdom
42,626
z-ai/glm-4.7
moonshotai/kimi-k2.5
communication_style_and_presentation
tie
61
Asian
Democrat
United States
13,185
google/gemini-2.0-flash-001
google/gemini-2.5-flash
trust_ethics_and_safety
tie
31
Mixed
Independent
United States
37,006
anthropic/claude-opus-4.5
cohere/command-r7b-12-2024
core_task_performance_and_reasoning
tie
53
White
null
United Kingdom
31,373
openai/gpt-4o
mistralai/mistral-large-3
communication_style_and_presentation
B
47
White
null
United Kingdom
34,665
meta-llama/llama-4-maverick
mistralai/mistral-large-3
trust_ethics_and_safety
B
55
Black
Democrat
United States
29,934
google/gemini-3-pro
openai/gpt-4o
core_task_performance_and_reasoning
A
66
White
Independent
United States
14,972
moonshotai/kimi-k2
google/gemini-2.5-pro
overall_winner
A
43
White
Republican
United States
37,169
google/gemini-3-pro
anthropic/claude-opus-4.5
communication_style_and_presentation
A
23
Black
Labour
United Kingdom
5,447
cohere/command-a
deepseek/deepseek-chat-v3-0324
communication_style_and_presentation
A
26
White
Liberal Democrats
United Kingdom
5,284
cohere/command-r7b-12-2024
google/gemini-2.5-flash
communication_style_and_presentation
B
59
White
Labour
United Kingdom
28,414
cohere/command-a
openai/gpt-4o
communication_style_and_presentation
A
71
White
Independent
United States
16,087
openai/o3
google/gemini-2.0-flash-001
communication_style_and_presentation
A
64
White
null
United States
53,669
x-ai/grok-4.20-beta
allenai/olmo-3.1-32b-instruct
communication_style_and_presentation
A
59
White
Liberal Democrats
United Kingdom
26,836
anthropic/claude-sonnet-4.5
cohere/command-r7b-12-2024
core_task_performance_and_reasoning
tie
76
White
Green Party
United Kingdom
9,543
meta-llama/llama-4-maverick
openai/o1-mini
communication_style_and_presentation
A
60
White
null
United Kingdom
52,517
qwen/qwen3-235b-a22b-2507
openai/gpt-5.4
overall_winner
A
58
White
Liberal Democrats
United Kingdom
31,947
mistralai/mistral-large-3
deepseek/deepseek-v3.2
core_task_performance_and_reasoning
B
51
White
Conservative
United Kingdom
52,912
moonshotai/kimi-k2.5
google/gemini-3.1-pro-preview
trust_ethics_and_safety
tie
63
White
Conservative
United Kingdom
7,498
mistralai/mistral-nemo
google/gemini-2.5-flash
communication_style_and_presentation
B
36
Latino/Hispanic
Independent
United States
13,327
x-ai/grok-3
openai/o3
core_task_performance_and_reasoning
tie
23
Asian
Conservative
United Kingdom
36,576
deepseek/deepseek-v3.2
x-ai/grok-3
trust_ethics_and_safety
tie
50
Mixed
Labour
United Kingdom
41,078
z-ai/glm-4.7
openai/gpt-5.2-chat
trust_ethics_and_safety
B
50
White
Democrat
United States
53,590
allenai/olmo-3.1-32b-instruct
anthropic/claude-opus-4.5
trust_ethics_and_safety
tie
23
Asian
null
United Kingdom
2,492
meta-llama/llama-3.3-70b-instruct
openai/gpt-4o
overall_winner
A
21
Latino/Hispanic
null
United States
18,716
openai/o1-mini
meta-llama/llama-3.3-70b-instruct
interaction_fluidity_and_adaptiveness
tie
52
Asian
null
United Kingdom
End of preview. Expand in Data Studio

HUMAINE: Human-AI Interaction Evaluation Dataset

Dataset Description

Dataset Summary

The HUMAINE dataset contains human evaluations of AI model interactions across diverse demographic groups and conversation contexts. This dataset powers the HUMAINE Leaderboard, providing insights into how different AI models perform across various user populations and use cases.

The dataset consists of two main components:

  • Feedback Comparisons: Pairwise model comparisons across multiple evaluation metrics
  • Conversations Metadata: Conversations with task complexity, achievement, and engagement scores

Note: There may be a slight discrepancy between the numbers in this dataset and the leaderboard app due to changes in consent related to data release and the post-processing steps involved in preparing this dataset.

Supported Tasks

  • Model performance evaluation
  • Demographic bias analysis
  • Preference learning
  • Human-AI interaction research
  • Conversational AI benchmarking

Dataset Structure

Data Files

The dataset contains two CSV files:

  1. feedback_dataset.csv

    • Pairwise comparisons between different AI models
    • Includes demographic information and preference choices
  2. conversations_metadata_dataset.csv

    • Metadata about individual conversations between users and AI models
    • Includes task types, domains, and performance scores

Data Fields

Feedback Comparisons

  • conversation_id: Unique identifier linking to conversation metadata
  • model_a: First model in the comparison
  • model_b: Second model in the comparison
  • metric: Evaluation metric (overall_winner, trust_ethics_and_safety, core_task_performance_and_reasoning, interaction_fluidity_and_adaptiveness)
  • choice: User's choice (A, B, or tie)
  • age: Age of the evaluator
  • ethnic_group: Ethnic group of the evaluator
  • political_affilation: Political affiliation of the evaluator
  • country_of_residence: Country of residence of the evaluator

Conversations Metadata

  • conversation_id: Unique identifier for the conversation
  • model_name: Name of the AI model used
  • task_type: Type of task (information_seeking, technical_assistance, etc.)
  • domain: Domain of the conversation (health_medical, technology, travel, etc.)
  • task_complexity_score: Complexity rating (1-5)
  • goal_achievement_score: How well the goal was achieved (1-5)
  • user_engagement_score: User engagement level (1-5)
  • total_messages: Total number of messages in the conversation

Usage

This dataset contains two CSV files that can be joined on the conversation_id field:

  • feedback_dataset.csv: Pairwise model comparisons with demographic information (primary dataset)
  • conversations_metadata_dataset.csv: Metadata about each conversation

Both files are included in this single dataset repository and can be accessed using HuggingFace's dataset loading utilities.

Dataset Creation

Curation Rationale

This dataset was created to address the lack of diverse, demographically-aware evaluation data for AI models. It captures real-world human preferences and interactions across different population groups, enabling more inclusive AI development.

Source Data

Data was collected through structured human evaluation tasks where participants:

  1. Engaged in conversations with various AI models
  2. Provided pairwise comparisons between model outputs
  3. Rated conversations on multiple quality dimensions (metrics)

Annotations

All annotations were provided by human evaluators through the Prolific platform, ensuring demographic diversity and high-quality feedback.

Personal and Sensitive Information

The dataset contains aggregated demographic information (age groups, ethnic groups, political affiliations, countries) but no personally identifiable information. All data has been anonymized and aggregated to protect participant privacy.

Considerations for Using the Data

Social Impact

This dataset aims to promote more inclusive AI development by highlighting performance differences across demographic groups. It should be used to improve AI systems' fairness and effectiveness for all users.

Discussion of Biases

While efforts were made to ensure demographic diversity, the dataset may still contain biases related to:

  • Geographic representation (primarily US and UK participants)
  • Self-selection bias in participant recruitment
  • Cultural and linguistic factors affecting evaluation criteria

Other Known Limitations

  • Limited to English-language interactions
  • Demographic categories are self-reported
  • Temporal bias (models evaluated at specific points in time)

Additional Information

Dataset Curators

This dataset was curated by the Prolific AI team as part of the HUMAINE (Human-AI Interaction Evaluation) project.

Licensing Information

This dataset is released under the MIT License.

Citation Information

@dataset{humaine2025,
  title={HUMAINE: Human-AI Interaction Evaluation Dataset},
  author={Prolific AI Team},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/ProlificAI/humaine-evaluation-dataset}
}

Contributions

Thanks to all the human evaluators who contributed their feedback to this project!

Contact

For questions or feedback about this dataset, please visit the HUMAINE Leaderboard or contact the Prolific AI team.

Downloads last month
81