The research workspace

For universities and research groups: studies with coded participants, EDA memory tests from the published methods, classification and machine learning, and every number exported.

What a study holds

A study has two groups -- the participants who know the tested details and those who do not -- because every classification here asks how well a test tells them apart. Participants are codes (P007), with a group, and an age and sex if the design matches on them. No names, no accounts: a participant takes a test through a link opened on the researcher's computer, with the PsyCalm sensor on two fingertips.

The methods

Standard CIT with EDA

A question, then its answers one at a time, each for five seconds with a blank screen of ten to fifteen between, while skin conductance is recorded (Lykken, 1959).

Interactive slide selection (ISS)

Irrelevant slides of one second each, with nothing between, until the signal has fallen for three seconds; only then the critical slide, and irrelevant ones again until the response is over. The relevant slide is never shown on a rising signal, and the test is several times shorter.

Conditioning to true content (CtTC)

First questions whose true answer is known -- the participant's name, their year of birth -- with an unpleasant stimulus after every true statement. The participant learns to respond to whatever is true; then the test questions, with the conditioning repeated against extinction.

CtTC with slides chosen by ISS

The two together: the conditioning and the test presented by the interactive selection, each true statement shown only when the signal has settled.

The questionnaires and timed tasks of the Testing section are there for participants too.

What comes out

  • Every response in the latency window of 1-4 s after a slide: latency, amplitude, slope and rise time, absolute and relative to the participant's largest.
  • Each participant classified by Lykken's scoring and by Z-scores, and each method's sensitivity, specificity, precision, accuracy and ROC-AUC across the study.
  • Machine-learning classifiers (MLP, LDA, random forest, logistic regression, k-nearest neighbours, decision tree) with leave-one-out cross-validation, with and without the CIT-dataset sorting algorithm.
  • CSV files of every slide, every participant, the datasets and the raw signal, for R, Python or SPSS.

The unconditioned stimulus

The published studies used a short electric pulse whose strength each participant set. The PsyCalm only measures; it gives no stimulation. The workspace offers an unpleasant sound in headphones instead, at a level the participant sets before the test, as the literature proposes where electric stimulation is not wanted -- or a trigger pulse on the audio output for a stimulator of the researcher's own, which then sets the current.

The papers behind it

  • Randjelovic, V. (2023). Interactive slide selection algorithm and machine learning in psychophysiological memory testing. Physiological Measurement, 44, 025004.
  • Randjelovic, V. (2024). Conditioning to true content and artificial intelligence in psychophysiological intention recognition. International Journal of Psychophysiology, 197, 112296.
  • Randjelovic, V. (2024). Doctoral dissertation, Singidunum University, Belgrade.

Their results describe their own laboratory samples and conditions. This implementation follows their methods; it has not been validated as a whole, and its classifications describe only the study they come from. A memory test is not a lie detector.

Getting access

The workspace comes with the research plan. A study can be set up as soon as the plan is active; giving participants a test waits until staff have checked your institution.

Subscribe to PsyEDA PRO Sign in