Skip to main content

Table 6 Relative Frequencies and Results of a Wilcoxon Test for Paired Samples Between the AI and Human Ratings (N = 1080)

From: Characteristics of learning tasks in accounting textbooks: an AI assisted analysis

 

Human coding

AI coding

   

Construct, category level

0

1

2

0

1

2

Method

z-Value

p-value

Effect size r

Process orientation

85.9%

9.8%

4.3%

60.1%

27.2%

12.7%

NN

− 15.6

0

− 0.475

Description of social processes

92.0%

6.7%

1.3%

81.0%

16.7%

2.3%

NN

− 9.27

0

− 0.282

Description of goods and services

92.3%

7.7%

 

86.6%

13.4%

 

NN

− 6.54

0

− 0.199

Description of cash flows

90.1%

9.9%

 

83.5%

16.5%

 

NN

− 7.89

0

− 0.240

Document orientation

94.5%

0.0%

5.5%

94.7%

0.0%

5.3%

RF

1.41

.346

0.043

Identification of problems

53.9%

45.1%

1.0%

61.5%

37.6%

0.9%

RF

8.9

0

0.271

Obtaining missing information

69.4%

19.3%

11.3%

55.9%

27.5%

16.6%

NN

− 8.29

0

− 0.252

Assessing the relevance of information

98.4%

1.6%

 

100%

0.0%

 

NN

4.12

0

0.125

Translate

70.5%

29.5%

 

79.3%

20.7%

 

RF

9.75

0

0.297

Operate

38.0%

62.0%

 

43.2%

56.8%

 

RF

7.18

0

0.218

Interpret

79.4%

20.6%

 

74.3%

25.7%

 

NN

− 4.04

0

− 0.123

Interpret: liquidity aspects

97.2%

2.8%

 

93.3%

6.7%

 

NN

− 6.19

0

− 0.188

Interpret: profitability aspects

91.3%

8.7%

 

80.0%

20.0%

 

NN

− 10

0

− 0.304

Validate

99.8%

0.2%

 

99.8%

0.0%

 

RF

–/–

–/–

–/–

  1. NN  Neural Net, RF Random Forest