Introduction
Literature review
Personality
Personality modeling in computer-based learning
Proposed framework for modeling the learner’s personality
CAG game
Personality dimension | Game traces | Definition in the game |
---|---|---|
Extraversion | Time (T) | The total amount of time that the learner spends in reading the story. |
Confidence (C) | The use of the option “your score will be seen by all your friends” to verify if the learner will check this option. | |
Color (CR) | The color of the chosen clothes for the game character. | |
Accessed Areas (AA) | The places that the learner visited while exploring the game environment. | |
Risk (R) | The paths that the learner decides to follow (risky or safe paths). | |
Number of friends (NF) | Number of friends that the learner will make in the game. | |
Feeling (F) | The positive or negative feelings that the learner has toward making conversation with NPCs. This can be done by freely choosing to accept or reject to start this conversation. | |
Gregariousness (G) | Learner’s preference for accompanying NPCs for a tour in the game. | |
Openness | Time (T) | The time that the learner spends in reading the story. |
Score (SCR) | The accumulated score that the learner earned during the learning-playing process. | |
Accessed Areas (AA) | The places that the learner visited while exploring the game environment. |
Learning analytics system
Dimension | Classification rules |
---|---|
Extraversion | If T is high AND C is high AND AA is sociable AND R is risk taker AND NF is high AND CR is warm AND G is yes AND F is positive THEN high extraversion |
If T is medium AND C is medium AND AA is sociable AND R is not risk taker AND NF is high AND CR is cool AND G is yes AND F is negative THEN balanced extraversion | |
If T is low AND C is low AND AA is not sociable AND R is not risk taker AND NF is low AND CR is cool AND G is no AND F is negative THEN low extraversion | |
Openness | If T is high AND SCR is high AND AA is sociable THEN high openness |
If T is medium AND SCR is low AND AA is not sociable THEN balanced openness | |
If T is low AND SCR is low AND AA is not sociable THEN low openness |
Method
Participants
Procedure
Instruments
Results
Accuracy results
Personality dimension | Personality Dimension Level | Number of learners (Using the BFI) | Number of wrong results (Using our system) | Number of correct results (Using our system) | Accuracy level |
---|---|---|---|---|---|
Extraversion | High Extraversion | 23 | 3 | 20 | 79.41% |
Balanced Extraversion | 2 | 0 | 2 | ||
Low Extraversion | 9 | 4 | 5 | ||
Openness | High Openness | 24 | 6 | 18 | 70.58% |
Balanced Openness | 0 | 0 | 0 | ||
Low Openness | 10 | 4 | 6 |
Agreement degree results
Personality dimension | Value | Asymp. Std. Error | Approx. T | Approx. Sig. |
---|---|---|---|---|
Extraversion | .651 | .122 | 4.425 | .000 |
Openness | .57 | .245 | 4.567 | .000 |
Technology acceptance results
TAM variables | Inter-items | Cronbach’s alpha | Mean | Median |
---|---|---|---|---|
EOU | 3 | .82 | 1.29 | 1 |
U | 3 | .75 | 1.15 | 1 |
ATT | 4 | .89 | 1.33 | 1 |
INT | 3 | .91 | 1.37 | 1 |
Discussion
Systems | Learning | Fun | Traces | Personality Model | Data Analysis Method | Accuracy |
---|---|---|---|---|---|---|
Virtual Personality Assessment Lab (Bunian et al. 2018) | – | + | Gaming behavior | FFM | Hidden Markov Models (HMM), Baum-Welch algorithm | From 54.1% to 59.1% |
Psyops (Tekofsky et al. 2013) | – | + | Gaming behavior | FFM | Not mentioned | Not mentioned |
Handwriting (Chen and Lin 2017) | – | – | Hand-writing | Not mentioned | Support Vector Machine, k-Nearest Neighbour, AdaBoost and Artificial Neural Network | From 62.5% to 83.9% |
MOOC (Chen et al. 2016) | + | – | Learning | FFM | Gaussian Process and Random Forest | Not mentioned |
Facebook (Buettner 2017). | – | – | Social network | FFM | Generalized linear modeling | From 62% to 71% |
Twitter (Golbeck et al. 2011) | – | – | Social network | FFM | ZeroR and Gaussian Processes | Not mentioned |
Electronically Activated Recorder (Mairesse et al. 2007) | – | – | Speech | FFM | Naive Bayes, AdaboostM1 and Support vector machines | From 51.45% to 62.52% |
Smart phones (Chittaranjan et al. 2011) | – | – | Smart phone | FFM | SVM and C4.5 classifiers | From 59.8% to 75.9% |
E-learning system (Ghorbani and Montazer 2015) | + | – | Learning | FFM | Fuzzy logic | From 78% to 97% |
Wearable sensors (Olguın et al. 2009) | – | – | Sensors | FFM | Accelerometer signal, IR transmissions, RSSI (radio signal strength indicator), | Not mentioned |
Our framework (CAG + LA system) | + | + | Gaming behavior | FFM | Naïve Bayes classifier | From 70.58% to 79.41% |