eye_movements_personality/README.md

76 lines
6.5 KiB
Markdown
Raw Normal View History

2018-04-10 22:33:06 +02:00
# Eye movements during everyday behavior predict personality traits
*Sabrina Hoppe, Tobias Loetscher, Stephanie Morey and Andreas Bulling*
2018-05-05 21:25:39 +02:00
This repository provides all data and code used for the publication [in Frontiers in Human Neuroscience](https://dx.doi.org/10.3389/fnhum.2018.00105).
2018-04-10 22:33:06 +02:00
## Dataset
* Gaze data recorded at 60Hz from 42 participants is stored in `data/ParticipantXX`.
For each participant there are three files:
1. `events.csv` is a list of gaze events as provided by the SMI eye tracker software.
The list contains saccades, fixations and blinks but only the blink information was used in the code.
2. `gaze_positions.csv` is a table with three columns: time in seconds, x gaze coordinate and y gaze coordinate. The x and y coordinates describe the participants' gaze direction normalised to the range from 0 to 1.
3. `pupil_diameter.csv` is another table with three columns: time in seconds, diameter of the right eye and diameter of the left eye. The diameter values are absolute gaze estimates in mm.
All files are of the same length and each row corresponds to one data sample. That is, the n-th row in all three files belongs to the same point in time.
* Ground truth personality scores from the respective questionnaires, participant age and sex (1: male, 2: female) can be found in `info/personality_sex_age.csv`.
* Personality score ranges that were obtained by binning the questionnaire scores are provided in `info/binned_personality.csv`.
* Timestamps indicating the times when participants entered and left the shop are given in `info/annotation.csv` in seconds.
2018-05-05 21:25:39 +02:00
## Code
reproducing the paper results step by step:
1. __Extract features from raw gaze data__:
2018-05-05 21:26:56 +02:00
`python 00_compute_features.py` to compute gaze features for all participants
2018-05-05 21:25:39 +02:00
Once extracted, the features are stored in `features/ParticipantXX/window_features_YY.npy` where XX is the participant number and YY the length of the sliding window in seconds.
2018-05-05 22:05:03 +02:00
2. __Train random forest classifiers__
`./01 train_classifiers.sh` to reproduce the evaluation setting described in the paper in which each classifier was trained 100 times.
`./02_train_specialized_classifiers.sh` to train specialized classifiers on parts of the data (specifically on data from inside the shop or on the way).
2018-05-05 21:25:39 +02:00
2018-05-05 22:05:03 +02:00
If the scripts cannot be executed, you might not have the right access permissions to do so. On Linux, you can try `chmod +x 01_train_classifiers.sh`,`chmod +x 02_train_specialized_classifiers.sh` and `chmod +x 03_label_permutation_test.sh` (see below for when/how to use the last script).
In case you want to call the script differently, e.g. to speed-up the computation or try with different parameters, you can pass the following arguments to `classifiers.train_classifier`:
`-t` trait index between 0 and 6
`-l` lowest number of repetitions, e.g. 0
`-m` max number of repetitions, e.g. 100
`-a` using partial data only: 0 (all data), 1 (way data), 2(shop data)
In case of performance issues, it might be useful to check `_conf.py` and change `max_n_jobs` to restrict the number of jobs (i.e. threads) running in parallel.
The results will be saved in `results/A0` for all data, `results/A1` for way data only and `results/A2` for data inside a shop. Each file is named `TTT_XXX.npz`, where TTT is the abbreviation of the personality trait (`O`,`C`,`E`,`A`,`N` for the Big Five and `CEI` or `PCS` for the two curiosity measures). XXX enumerates the classifiers (remember that we always train 100 classifiers for evaluation because there is some randomness involved in the training process).
3. __Evaluate Baselines__
* To train a classifier that always predicts the most frequent personality score range from its current training set, please execute `python 03_train_baseline.py`
* To train classifiers on permuted labels, i.e. perform the so-called label permutation test, please execute `./04_label_permutation_test.sh`
2018-05-05 21:25:39 +02:00
2018-05-05 21:26:56 +02:00
2018-04-10 22:33:06 +02:00
## Citation
If you want to cite this project, please use the following Bibtex format:
```
@article{hoppe18_fhns,
title = {Eye Movements During Everyday Behavior Predict Personality Traits},
author = {Sabrina Hoppe and Tobias Loetscher and Stephanie Morey and Andreas Bulling},
2018-05-05 20:55:03 +02:00
url = {https://perceptual.mpi-inf.mpg.de/files/2018/04/hoppe18_fhns.pdf
https://github.molgen.mpg.de/sabrina-hoppe/everyday-eye-movements-predict-personality
https://www.newscientist.com/article/2167850-ai-can-predict-your-personality-just-by-how-your-eyes-move/
http://www.dailymail.co.uk/sciencetech/article-5686817/An-incredible-mind-reading-AI-predict-personality-just-studying-eyes-move.html
https://www.digitaltrends.com/cool-tech/ai-personality-eye-movement/
http://www.newsweek.com/artificial-intelligence-algorithm-can-work-out-your-personality-simply-909752
https://www.12news.com/video/syndication/veuer/new-ai-can-predict-personality-from-eye-movements/602-8116328
https://www.usatoday.com/videos/tech/news/2018/05/03/new-ai-can-predict-personality-eye-movements/34526179/},
2018-04-10 22:33:06 +02:00
doi = {10.3389/fnhum.2018.00105},
year = {2018},
date = {2018-03-05},
journal = {Frontiers in Human Neuroscience},
volume = {12},
2018-05-05 20:55:03 +02:00
pages = {105:1-105:8},
2018-04-10 22:33:06 +02:00
abstract = {Besides allowing us to perceive our surroundings, eye movements are also a window into our mind and a rich source of information on who we are, how we feel, and what we do. Here we show that eye movements during an everyday task predict aspects of our personality. We tracked eye movements of 42 participants while they ran an errand on a university campus and subsequently assessed their personality traits using well-established questionnaires. Using a state-of-the-art machine learning method and a rich set of features encoding different eye movement characteristics, we were able to reliably predict four of the Big Five personality traits (neuroticism, extraversion, agreeableness, conscientiousness) as well as perceptual curiosity only from eye movements. Further analysis revealed new relations between previously neglected eye movement characteristics and personality. Our findings demonstrate a considerable influence of personality on everyday eye movement control, thereby complementing earlier studies in laboratory settings. Improving automatic recognition and interpretation of human social signals is an important endeavor, enabling innovative design of humancomputer systems capable of sensing spontaneous natural user behavior to facilitate efficient interaction and personalization.},
keywords = {},
2018-05-05 20:55:03 +02:00
pubstate = {published},
2018-04-10 22:33:06 +02:00
tppubtype = {article}
}
```