Table 3 The assessment of the abstract by four different methods other than academicians.
Variable | GPT-2 output detector (n) | ||||||
|---|---|---|---|---|---|---|---|
Abstract type | Low fake | Moderate fake | High fake | Very high fake | Pearson Chi-square * | P-value^ | Phi value |
Original abstract | 66 | 7 | 2 | 5 | 7.281 | 0.063 | 0.213 |
AI abstract | 53 | 12 | 9 | 6 | |||
Variable | Writefull GPT detector (n) | ||||||
Abstract type | Entirely human | Mostly human made | Partly by AI | Entirely by AI | Pearson Chi square | P-value | Phi value |
Original abstract | 62 | 2 | 5 | 11 | 18.705 | < 0.001 | 0.342 |
AI abstract | 38 | 13 | 14 | 15 | |||
Variable | GPTZero detector (n) | ||||||
Abstract type | Low fake | Moderate fake | High fake | Very high fake | Pearson Chi-square | P-value^ | Phi value |
Original abstract | 80 | 0 | 0 | 0 | 144.762 | < 0.001 | 0.951 |
AI abstract | 4 | 11 | 13 | 53 | |||
Variable | Similarity outcome (n) | ||||||
Abstract type | Low similarity | Moderate similarity | High similarity | Very high similarity | Pearson Chi square | P-value | Phi value |
Original abstract | 0 | 0 | 0 | 80 | 144.762 | < 0.001 | 0.951 |
AI abstract | 23 | 42 | 11 | 4 | |||
