# finetuning_data.RDS Every 10th training step, generation was prompted for three varying temperatures (0.7, 0.9, 1.1) - `id` id of the current training step. - `timestamp` timestamp of the current training step. - `duration` duration for item generation and training at current training step. - `training_step` training step with sample item generation (every 10th). - `loss` cross-entropy loss at training step. # training_data.RDS Training data adopted from Goldberg, L. R. (1999). A broad-bandwidth, public domain, personality inventory measuring the lower-level facets of several five-factor models. Personality Psychology in Europe, 7(1), 7–28. Goldberg, L. R., Johnson, J. A., Eber, H. W., Hogan, R., Ashton, M. C., Cloninger, C. R., & Gough, H. G. (2006). The international personality item pool and the future of public-domain personality measures. Journal of Research in Personality, 40(1), 84–96. `item` item stem `construct` construct label # generation_data.RDS Data from prompted item generation, after fine-tuning was completed. - `construct_{n}`: The nth construct the item stem is associated with. - `stem_1`: The item stem. - `similarity`: The Levenshtein distance to the most similar item in the training data. - `overfit`: `1` If the item exceeded a `similarity` of `.90`. - `in_training`: `TRUE` if the item was present in the training data. # content_validity.RDS Data from content validity ratings. - `id`: item id. - `rater_01`: `1` If the item was endorsed for content validity by rater 1. - `rater_02`: `1` If the item was endorsed for content validity by rater 2. - `construct_1`: The construct label associated with the item. - `stem_1`: The item stem. - `similarity`: The Levenshtein distance to the most similar item in the training data. - `overfit`: `1` If the item exceeded a `similarity` of `.90`. - `matches`: The item stems of the most similar items in the training data. - `content_validity`: `TRUE` if both raters endorsed the item for content validity. - `in_survey`: `TRUE` if the item was randomly assigned to be part of the online survey. - `in_survey_name`: variable name in `survey_data.RDS`. # survey_data.RDS Data from the online survey. - `id`: The participant ID. - `timestamp`: The timestamp of participation. - `subject_code`: The hased subject id. - `completed`: `TRUE` if the questionnaire was completed by the participant. - `strict_bogus`: The number of bogus points as an indicator of careless responding. - `time_rsi`: The relative speed index (Leiner, 2019). - `age`: The age of the participant. - `sex`: The biological sex of the participant (see factor labels). The following variables describe human- and machine-authored items, wheras `{n}` is the item number (1-5). - `{prefix}{n}_machine`: Machine-authored items (AIG). - `{prefix}{n}_human`: Human-authored items from the BFI-dataset in the R psych-package (version 2.0.9; Revelle, 2020, based on Goldberg, 1999). - `ope{n}_{suffix}`: Items of the scale **Openness to experience**. - `con{n}_{suffix}`: Items of the scale **Conscientiousness**. - `ext{n}_{suffix}`: Items of the scale **Extraversion**. - `agr{n}_{suffix}`: Items of the scale **Agreeableness**. - `neu{n}_{suffix}`: Items of the scale **Neuroticism**. - `ben{n}_{suffix}`: Items of the scale **Benevolenc**`. - `ega{n}_{suffix}`: Items of the scale **Egalitarianis**`. - `ego{n}_{suffix}`: Items of the scale **Egois**`. - `jov{n}_{suffix}`: Items of the scale **Jovialit**`. - `pes{n}_{suffix}`: Items of the scale **Pessimis**`. Leiner, D. J. (2019). Too fast, too straight, too weird: Non-reactive indicators for meaningless data in internet surveys. Survey Research Methods, 229-248 Pages. https://doi.org/10.18148/SRM/2019.V13I3.7403 # boot_untrained_scales.RDS Data to replicate bootstrapped results for omega-coefficients in Table 2. # boot_trained_scales.RDS Data to replicate bootstrapped results for omega-coefficients in Table 1. # power_simulation.RDS Data to replicate the power simulation described in paper. [\[back to main repository\]][11] [11]: https://osf.io/3bh7d/