As industries progress toward integrating more complex technologies within Industry 4.0 frameworks, ensuring work instructions that balance cognitive load and performance is increasingly critical, especially under the human-centric principles of the 5th industrial revolution. Drawing on Cognitive Load Theory (CLT), this study compares two instructional methods-visual-based and code-based-to determine whether cognitive overload can be reduced without compromising task outcomes in a controlled, assembly-like scenario derived from industrial tasks. We recruited 30 participants from the academic field (students and researchers), who completed assembly tasks under both visual-based and code-based instructions. Cognitive load was measured objectively by (Galvanic Skin Response, Heart Rate Variability, and hand motion acceleration) and subjectively through (NASA Task Load Index, short Dundee Stress State Questionnaire). Operational efficiency was assessed via task completion time (TCT), number of task repetitions (NTR), and assembly precision based on the standard deviation. The findings demonstrated that visual-based instructions significantly reduced cognitive load with a p - value < 0.001. It also showed an improvement in two of the performance metrics during the use of visual-based instructions for the TCT and NTR with p - value < 0.001. However, although code-based instructions increased cognitive load, they showed better assembly precision with a p - value < 0.001. These results suggest that while simple and direct instructions facilitate task execution and reduce cognitive loads, deep thinking approaches may still hold value for tasks requiring high precision.
Ethical approval number: KEB_MK_FIT_2024_01
What the experiment is
A within-subject, counterbalanced study (N = 30) where participants assembled “Make ’N’ Break Extreme” patterns twice—once with visual-based step images (low extraneous load) and once with code-based alphanumeric/color cues (higher interpretive demand). Before any task, a 3-minute baseline was recorded. After each session, participants completed NASA-TLX and the short DSSQ (engagement, distress, worry). Objective signals were collected with Shimmer3 (GSR, PPG→HRV) and a MetaMotion wrist accelerometer on the dominant hand; assembly precision was computed from video using ArUco markers (lower SD = better precision). Each session targeted ~5 min and ≥3 repetitions; order was counterbalanced (half started Visual, half Code).
Excel file (“Extracted data – with performance.xlsx”)
Column prefixes:
V* → metrics from the Visual instruction session; C* → Code instruction session.
F – Sessions: sequence for each participant: V_C = Visual then Code; C_V = Code then Visual.
G–T – NASA_TLX: subscales (mental, physical, temporal demand, performance, effort, frustration) and any composite/weights used in RTLX.
U–AC – DSSQ: short-form DSSQ states across the required timepoints (pre / post-visual / post-code), typically Engagement, Distress, Worry.
AD–AE – Precision: assembly precision summarized as standard deviation of Euclidean distances between block centers (video-tracked); lower values mean better precision. (Precision methodology per paper.)
Physiological raw-data package (“Physiological Raw Data/”)
Top-level grouping
G1/ and G2/ split the cohort into two counterbalanced groups
Within each group
Sub1/ and Sub2/ are the sub-groups used to balance which pattern sets (1&2 vs. 3&4) are paired with each instruction order.
Each participant has a numbered folder: 1/, 2/, 3/, …
Inside each participant folder
Session & activity timeline (*.csv): time-stamped report of session boundaries and key events.
Acceleration (dominant_hand_accel.csv): tri-axial wrist accelerometer log for the dominant hand.
Shimmer3 physiological data (MultiSession/): three CSVs:
Baseline (Session1),
Session 2, Session 3 (Visual or Code depending on that participant’s Sessions order in the Extracted data – with performance.xlsx). Signals: GSR and PPG (for HRV extraction). In the study, Shimmer3 sampling was set to 250 Hz.