--- license: apache-2.0 dataset_info: features: - name: case dtype: int64 - name: sex dtype: int64 - name: age dtype: int64 - name: sec dtype: int64 - name: min dtype: int64 - name: hour dtype: int64 - name: day dtype: int64 - name: month dtype: int64 - name: year dtype: int64 - name: country dtype: int64 - name: i1 dtype: int64 - name: i2 dtype: int64 - name: i3 dtype: int64 - name: i4 dtype: int64 - name: i5 dtype: int64 - name: i6 dtype: int64 - name: i7 dtype: int64 - name: i8 dtype: int64 - name: i9 dtype: int64 - name: i10 dtype: int64 - name: i11 dtype: int64 - name: i12 dtype: int64 - name: i13 dtype: int64 - name: i14 dtype: int64 - name: i15 dtype: int64 - name: i16 dtype: int64 - name: i17 dtype: int64 - name: i18 dtype: int64 - name: i19 dtype: int64 - name: i20 dtype: int64 - name: i21 dtype: int64 - name: i22 dtype: int64 - name: i23 dtype: int64 - name: i24 dtype: int64 - name: i25 dtype: int64 - name: i26 dtype: int64 - name: i27 dtype: int64 - name: i28 dtype: int64 - name: i29 dtype: int64 - name: i30 dtype: int64 - name: i31 dtype: int64 - name: i32 dtype: int64 - name: i33 dtype: int64 - name: i34 dtype: int64 - name: i35 dtype: int64 - name: i36 dtype: int64 - name: i37 dtype: int64 - name: i38 dtype: int64 - name: i39 dtype: int64 - name: i40 dtype: int64 - name: i41 dtype: int64 - name: i42 dtype: int64 - name: i43 dtype: int64 - name: i44 dtype: int64 - name: i45 dtype: int64 - name: i46 dtype: int64 - name: i47 dtype: int64 - name: i48 dtype: int64 - name: i49 dtype: int64 - name: i50 dtype: int64 - name: i51 dtype: int64 - name: i52 dtype: int64 - name: i53 dtype: int64 - name: i54 dtype: int64 - name: i55 dtype: int64 - name: i56 dtype: int64 - name: i57 dtype: int64 - name: i58 dtype: int64 - name: i59 dtype: int64 - name: i60 dtype: int64 - name: i61 dtype: int64 - name: i62 dtype: int64 - name: i63 dtype: int64 - name: i64 dtype: int64 - name: i65 dtype: int64 - name: i66 dtype: int64 - name: i67 dtype: int64 - name: i68 dtype: int64 - name: i69 dtype: int64 - name: i70 dtype: int64 - name: i71 dtype: int64 - name: i72 dtype: int64 - name: i73 dtype: int64 - name: i74 dtype: int64 - name: i75 dtype: int64 - name: i76 dtype: int64 - name: i77 dtype: int64 - name: i78 dtype: int64 - name: i79 dtype: int64 - name: i80 dtype: int64 - name: i81 dtype: int64 - name: i82 dtype: int64 - name: i83 dtype: int64 - name: i84 dtype: int64 - name: i85 dtype: int64 - name: i86 dtype: int64 - name: i87 dtype: int64 - name: i88 dtype: int64 - name: i89 dtype: int64 - name: i90 dtype: int64 - name: i91 dtype: int64 - name: i92 dtype: int64 - name: i93 dtype: int64 - name: i94 dtype: int64 - name: i95 dtype: int64 - name: i96 dtype: int64 - name: i97 dtype: int64 - name: i98 dtype: int64 - name: i99 dtype: int64 - name: i100 dtype: int64 - name: i101 dtype: int64 - name: i102 dtype: int64 - name: i103 dtype: int64 - name: i104 dtype: int64 - name: i105 dtype: int64 - name: i106 dtype: int64 - name: i107 dtype: int64 - name: i108 dtype: int64 - name: i109 dtype: int64 - name: i110 dtype: int64 - name: i111 dtype: int64 - name: i112 dtype: int64 - name: i113 dtype: int64 - name: i114 dtype: int64 - name: i115 dtype: int64 - name: i116 dtype: int64 - name: i117 dtype: int64 - name: i118 dtype: int64 - name: i119 dtype: int64 - name: i120 dtype: int64 splits: - name: train num_bytes: 643916000 num_examples: 619150 download_size: 448482872 dataset_size: 643916000 --- # Personality Alignment with Personality Inventories (PAPI) Dataset - 600K Version ## Overview The 600K Personality Alignment with Personality Inventories (PAPI) dataset is an expanded version of the original PAPI dataset, focusing solely on the IPIP-NEO-120 questionnaire. This large-scale collection of personality assessments is designed to provide a more comprehensive resource for research in personality psychology and AI alignment, with a particular emphasis on the shorter IPIP-NEO-120 format. ## Dataset Description ### Size and Structure - Total samples: Approximately 619150 - Each sample contains: - IPIP-NEO-120 questionnaire responses ### Question Format - Each question is associated with one of the five major personality dimensions - Questions correlate positively or negatively with specific behavioral patterns or preferences - Five response options ranging from "Very Accurate" to "Very Inaccurate" ### Demographic Information - Gender distribution: - Female: ~60% - Male: ~40% - Age range: 10 to 99 years - Average age: Approximately 25 years - Geographical diversity: Participants from various countries including the USA, UK, France, India, and China ## Data Collection - Source: International Personality Item Pool (IPIP) - Collection period: 1998 to 2019 - Method: Online surveys - Survey duration: Approximately 15 to 25 minutes per participant (estimated for IPIP-NEO-120) - Anonymity: All responses are anonymous ## Data Processing - The dataset includes all available IPIP-NEO-120 responses, providing a larger sample size for analysis - No specific test set has been extracted; researchers can create their own train/test splits as needed ## Evaluation Methods - Behavioral difference score can be used as an evaluation metric - Scoring function assigns values from 5 to 1 for responses ranging from "Very Accurate" to "Very Inaccurate" - Aligned Score can be calculated for each Big Five trait based on the IPIP-NEO-120 responses ## Ethical Considerations - Voluntary participation with no monetary compensation - Data anonymization: No personally identifiable information (PII) included - Compliance with data protection regulations (e.g., GDPR) - Adherence to ethical principles outlined in the Belmont Report ## Dataset Analysis Users of this dataset are encouraged to perform their own analysis, which may include: - Demographic distributions (age, gender, country) - Personality trait distributions for the Big Five traits - Correlations between different personality traits - Comparative studies with other personality assessment tools ## Usage and Permissions The 600K PAPI dataset is available for research purposes. Users are granted permission to use IPIP items, scales, and inventories for any purpose, commercial or non-commercial. ## Citation If you use this dataset in your research, please cite: ``` @misc{zhu2024personalityalignmentlargelanguage, title={Personality Alignment of Large Language Models}, author={Minjun Zhu and Linyi Yang and Yue Zhang}, year={2024}, eprint={2408.11779}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2408.11779}, } ``` ## Contact For questions or additional information about the PAPI dataset, please contact: zhuminjun@westlake.edu.cn ## Acknowledgments We thank the administrators of the International Personality Item Pool (IPIP) for granting permission to use their items, scales, and inventories. We also express our gratitude to all the volunteers who participated in this study, contributing to the advancement of personality research and AI alignment. ## Note on Dataset Differences This 600K version of the PAPI dataset differs from the original version in the following ways: 1. It contains only IPIP-NEO-120 questionnaire responses, not IPIP-NEO-300. 2. It includes a larger number of samples (approximately 600,000 vs. 307,313). 3. It does not have a pre-defined test set, allowing researchers more flexibility in data usage. Researchers should consider these differences when deciding which version of the dataset to use for their specific research needs.