Multimodal Dataset of a Serious Game about Climate Change

Overview The Multimodal Dataset of a Serious Game about Climate Change is designed to study the effect of dynamic color saturation on learning, player experience, and emotional and cognitive processes during a serious game about climate change. The dataset captures dynamic human interactions through electroencephalography (EEG), electrodermal activity (EDA), in-game behavioral events, and pre- and post-self-report questionnaires. This enables the study of cognitive and affective responses to educational gameplay from multiple complementary perspectives. The study employed a between-subjects design, consisting of a treatment group in which color saturation was increased during key learning and quiz sequences and a control group in which a consistent, slightly desaturated setting was maintained. Participants were unaware of their assigned condition. The dataset comprises 50 participants. The serious game, Island Voices, is publicly available on Itch.io. Experimental Manipulation (Treatment vs. Control) Color saturation was applied as an overlay filter to all visual elements with 3-second ease-in/ease-out transitions (high = 110%, returning to 65%). For the treatment group, saturation shifts were tied to specific learning-relevant gameplay events. Interacting with informative objects or NPCs (Information Object, Information NPC). Completing quizzes (Quiz). Making decisions (Decision). Tutorial. These are the same event categories recorded in the behavioral event files (see below), so the event timeline also serves as the timeline of saturation manipulations. Dataset Structure The dataset is organised in a BIDS-style (Brain Imaging Data Structure) layout: Root folder sub-* : Individual subject folders (e.g., sub-P001, sub-P002, sub-P003) participants.tsv : Participant demographics and summary measures (with participants.json data dictionary) self-report.tsv : Summarized self-report data Subject folders (sub-<subject_id>) EEG Data (eeg/): EEG/ECG recordings (.tsv) with metadata (.json) EDA Data (eda/): Electrodermal activity recordings (.tsv) with session-level metadata (.json) Behavioural Data (beh/): In-game event timelines (.json) Data Modalities and Channels 1. EEG Data Device: B-Alert X10 (Advanced Brain Monitoring), recorded via iMotions Channels: 9 monopolar EEG channels in a reduced International 10-20 layout (Fz, F3, F4, Cz, C3, C4, P3, P4, POz) plus 1 differential ECG channel; values in microvolts Sampling Rate: 256 Hz (16-bit resolution) Reference: Linked mastoids (online reference) Time base: A time column in seconds, aligned to a common game time and shifted so the baseline is negative (see Usage Notes). The TSV also retains the iMotions synchronization timestamp and stimulus/slide-event columns. File Example: sub-<subject>/eeg/sub-<subject>_eeg.tsv (+ _eeg.json) Note: ECG was recorded for completeness and may not be analysed in associated work. 2. Electrodermal Activity (EDA) Data Device: Mindfield eSense Skin Response sensor with the "Mindfield eSense Biofeedback" mobile app; two reusable finger-clip Ag/AgCl electrodes on the tips of the index and middle fingers (with electrode gel) Channels: Skin conductance in microsiemens (microsiemens) with a time column (second) Sampling Rate: ≈5 Hz (0.2 s intervals) Time base: second shifted so the first baseline sample = −32 s (see Usage Notes) File Example: sub-<subject>/eda/sub-<subject>_eda.tsv (+ _eda.json) 3. In-Game Behavioural Events Content: Onset/offset intervals (in seconds) for interactions with informative objects and NPCs, quizzes (1–5), decisions (1–2), and the tutorial together with the per-participant game-time alignment value (full_timeline_diff) Format: One JSON object per participant (events as arrays of [start, end] pairs) File Example: sub-<subject>/beh/sub-<subject>_events.json 4. Demographics and Self-Report Measures Stored per participant in participants.tsv (documented by participants.json): Demographics: participant_id, experimental group (treatment / control), gender, age, weekly video-game experience, and a climate-change attitude score (Christensen, 2015) Knowledge / learning gain: pre- and post-test scores (10-item single-choice knowledge test, max 25 points) and the normalized gain (Hake, 1998) Affect (PANAS, German adaptation (Breyer, 2016)): Positive Affect (PA) and Negative Affect (NA) subscale means, before and after play (5-point scale) Perceived mental effort (Paas, 1992): single item, 7-point scale Player experience (PXI, German version (Graf, 2022)): overall plus functional and psycho-social consequence subscale scores (33 items, 7-point scale from −3 to +3) Experimental Setup Participants played the game on a 24-inch Dell P2412H monitor at 1920 × 1080 (60 Hz), RGB output at 8-bit depth in the standard dynamic range colour space (≈81 % gamut), with RGB channels at 100 % and brightness/contrast at 75 %. Lighting was held constant across sessions (windows covered, room light on), and viewing distance was kept as constant as possible. A resting baseline preceded gameplay; physiological signals (EEG, EDA) were recorded continuously while in-game events were timestamped and later used to align the physiological recordings with the corresponding gameplay events. The target population was German-speaking adults (18+) from a range of backgrounds, with limited prior knowledge of climate change and sea level rise. Ethical review and approval were waived by the Research Ethics Committee of TH Köln – University of Applied Sciences (application number THK-2025-0012). Usage Notes Encoding & formats: UTF-8; data stored as TSV (signals/tables) and JSON (metadata/events). Time alignment: All time axes are in seconds and aligned to a common game time. EEG times are computed as time = Timestamp/1000 − full_timeline_diff − baseline, and EDA times are shifted by the same baseline. By convention the baseline is negative: the first baseline sample is at −32 s, and 0 s marks the start of gameplay. full_timeline_diff is provided per participant (originally mm.ss, i.e. minutes.seconds) and was used to align each recording to game time. Language: Recording instruments and questionnaire items are in German; some metadata fields in the JSON sidecars retain their original German labels. Instruments: PXI (Graf, 2022), PANAS German adaptation (Breyer, 2016), perceived mental effort (Paas, 1992), normalized learning gain (Hake, 1998). Citation: Please cite this dataset (and the related publication) appropriately.