Result
Result
After collecting survey responses from students at Goizueta Business School, I used SPSS and AMOS to explore the relationships between AI literacy, trust, perceived usefulness, and the intention to use AI. Here’s a quick look at what the data revealed.
Table 1: Reliability and validity test results(SPSS)
| Construct | CR | AVE | Cronbach's α (Range) | Overall Cronbach's α | χ²/df | PNFI | CFI | IFI | RFI | RMSEA |
|---|---|---|---|---|---|---|---|---|---|---|
| AI Literacy | 0.821 | 0.546 | 0.545 - 0.880 | 0.819 | 1.124 | 0.877 | 0.987 | 0.980 | 0.987 | 0.034 |
| Perceived Ease of Use | 0.828 | 0.557 | 0.550 - 0.889 | 0.808 | 1.697 | 0.787 | 0.991 | 0.990 | 0.991 | 0.045 |
| Perceived Usefulness | 0.860 | 0.606 | 0.736 - 0.814 | 0.854 | 1.568 | 0.812 | 0.949 | 0.950 | 0.949 | 0.035 |
| Perceived Credibility | 0.875 | 0.700 | 0.800 - 0.871 | 0.873 | - | - | - | - | - | - |
| Intention to Use | 0.864 | 0.615 | 0.750 - 0.822 | 0.864 | 1.476 | 0.824 | 0.995 | 0.995 | 0.985 | 0.070 |
Before I could analyze students’ AI behavior, I had to make sure that the survey itself was trustworthy. That’s where these numbers come in. They may look complicated, but here’s what they really mean:
CR (Composite Reliability): This tells us if a group of related questions are consistent with each other. Think of it like a band playing in sync. A score above 0.7 is good.
AVE (Average Variance Extracted): This checks whether the questions are really measuring what they’re supposed to. Above 0.5 means the items are on target.
Cronbach’s α (Alpha): A classic measure of how reliable the questions are. If people respond consistently to similar items, the α will be high. 0.7 is acceptable, 0.8+ is great.
Overall Cronbach’s α: This is the α for the entire survey — like asking “is the whole test reliable?
χ²/df: Lower is better. Under 5 is usually okay.
PNFI, CFI, IFI, RFI: These are different model fit indexes. They all range from 0 to 1. Above 0.9 is considered a good fit.
RMSEA (Root Mean Square Error of Approximation): This shows how much “error” there is in the model on average. Less than 0.08 is acceptable. Less than 0.05 is excellent.
This table tells us, Yes, the survey questions are reliable.
Table 2. Measurement model (AMOS)
This table checks if each survey question is actually measuring what it’s supposed to.
For example, to measure AI literacy, I asked several questions. This chart confirms whether
those questions really fit together and make sense. These numbers check how well my model fits the real data. Think of them as a “reality check” for the path diagram. Do the relationships I mapped out actually show up in the survey responses?
| Model | Type | X2 | Df | X2/Df | CFI | TLI | RMSEA | SRMR |
|---|---|---|---|---|---|---|---|---|
| 5-factor model | AIL,PEOU,PU,PC,IU | 224.672 | 142 | 1.582 | 0.917 | 0.900 | 0.077 | 0.067 |
| 4-factor model | AIL,PEOU+PU,PC,IU | 390.942 | 146 | 2.678 | 0.755 | 0.713 | 0.132 | 0.120 |
| 3-factor model | AIL,PEOU+PU+PC,IU | 531.698 | 149 | 3.568 | 0.617 | 0.560 | 0.163 | 0.148 |
| 2-factor model | AIL+PEOU+PU+PC,IU | 608.873 | 151 | 4.032 | 0.542 | 0.481 | 0.177 | 0.141 |
| 1-factor model | AIL+PEOU+PU+PC+IU | 625.321 | 152 | 4.114 | 0.526 | 0.467 | 0.179 | 0.142 |
AIL AI Literacy, PEOU Perceived Ease of Use, PU Perceived Usefulness, PC Perceived Credibility, IU Intention to Use
χ² and df (degrees of freedom): These work together. What we really care about is their ratio.
χ²/df: A smaller number is better. If it’s under 5, that means the model fit is acceptable.
CFI (Comparative Fit Index): Ranges from 0 to 1. The closer to 1, the better. A score above 0.9 is excellent.
TLI (Tucker–Lewis Index): Very similar to CFI. Also great if it’s above 0.9.
RMSEA (Root Mean Square Error of Approximation): Measures how much error is in the model. Lower = better. Below 0.08 is okay. Below 0.05 is great.
SRMR (Standardized Root Mean Square Residual): Tells us how far off the predictions were from real data. Anything under 0.08 is considered a good fit.
If these values hit the target range, the model is a good match with real student behavior.
Table 3: Frequency Analysis (SPSS)
This table shows a breakdown of who answered the survey: their gender, year in school, major, and how often they use AI tools. This helps us understand who the participants are and how balanced the sample is.
| Variable | Option | Frequency | Percentage (%) |
|---|---|---|---|
| Gender | Male | 58 | 59.2 |
| Female | 40 | 40.8 | |
| Freshman | 12 | 12.2 | |
| Sophomore | 12 | 12.2 | |
| Junior | 27 | 27.6 | |
| Senior | 43 | 43.9 | |
| Academic level | Graduate | 4 | 4.1 |
| Finance | 31 | 31.6 | |
| Accounting | 16 | 16.3 | |
| Marketing | 21 | 21.4 | |
| Management | 14 | 14.3 | |
| Major | Others | 16 | 16.3 |
| Frequently use | 35 | 35.7 | |
| Occasionally use | 60 | 61.2 | |
| Have you used AI tools to assist with learning or work? | Never used | 3 | 3.0 |
Table 4: Normality Test Analysis Results (SPSS)
This table checks whether the survey data follows a normal distribution, which is a basic requirement before running advanced analyses like structural modeling.
| Mean | Standard | Deviation | Skewness | Kurtosis |
|---|---|---|---|---|
| AI literacy | 3.314 | 0.734 | -0.436 | -0.859 |
| Perceived usefulness | 3.880 | 0.805 | -1.668 | 2.984 |
| Perceived ease of use | 3.850 | 0.916 | -1.386 | 1.122 |
| Perceived credibility | 3.214 | 1.069 | 0.230 | -1.427 |
| Intention to use | 4.000 | 0.762 | -1.403 | 1.744 |
* p<0.05 ** p<0.01
Mean: The average score for each item. On a 5-point scale, values around 3 to 4 are typical showing that students leaned toward agreement.
Standard Deviation: Measures how consistent the responses are. A smaller number = more consistent answers.
Skewness: Shows if the data leans to one side. Closer to 0 means more balanced. Values between -2 and +2 are usually okay.
Kurtosis: Tells us if responses are too clustered or too flat. Again, closer to 0 is better, and values between -7 and +7 are generally acceptable.
This table confirms that the data is clean, balanced, and ready for further analysis.
Table 5. Path analysis results (AMOS)
| Path | Unstandardized Path Coefficient | Standardized Path Coefficient | S.E. | C.R. | P |
|---|---|---|---|---|---|
| PEOU <--- AIL | 0.363 | 0.342 | 0.140 | 2.590 | 0.010 |
| PU <--- AIL | 0.701 | 0.423 | 0.212 | 3.310 | *** |
| PC <--- AIL | 0.887 | 0.442 | 0.252 | 3.516 | *** |
| IU <--- AIL | 0.478 | 0.321 | 0.172 | 2.780 | 0.005 |
| IU <--- PU | 0.324 | 0.360 | 0.086 | 3.758 | *** |
| IU <--- PEOU | 0.425 | 0.302 | 0.139 | 3.066 | 0.002 |
| IU <--- PC | 0.173 | 0.233 | 0.068 | 2.556 | 0.011 |
AIL AI Literacy, PEOU Perceived Ease of Use, PU Perceived Usefulness, PC Perceived Credibility, IU Intention to Use
This is one of the most important tables in the whole study. It shows which factors actually influence students’ intention to use AI, and how strong those effects are.
Unstandardized Path Coefficient: The raw effect — used for technical calculations.
Standardized Path Coefficient: This is the most important value. It tells us how strong the effect is on a consistent scale:
- 0.7 = very strong
- 0.3 = moderate
- Below 0.1 = weak or no effect
S.E. (Standard Error): Smaller means more reliable results.
C.R. (Critical Ratio): Like a t-score — used to test statistical significance.
P (p-value): This tells us if the effect is real.If p < 0.05, we can say the result is statistically significant.
This table tells us which factors matter, and how much they matter when it comes to students deciding whether they’ll use AI in their future jobs.
Table 6. Direct, indirect and total B coefficients (Bootstrap)
This table is all about mediating effects — whether certain variables act like a “bridge” between AI literacy and the intention to use AI.
| Total, direct, and indirect effect | Path Coefficien | Bias-Corrected | Percentile | ||
|---|---|---|---|---|---|
| Lower | Upper | Lower | Upper | ||
| AIL→IU Total effect | 0.679 | 0.461 | 0.815 | 0.476 | 0.821 |
| AIL→IU Direct effect | 0.321 | 0.029 | 0.58 | 0.002 | 0.565 |
| AIL→IU Indirect effect | 0.358 | 0.181 | 0.636 | 0.178 | 0.629 |
| AIL→PEOU→IU Individual indirect effect | 0.103 | 0.016 | 0.386 | 0.008 | 0.322 |
| AIL→PU→IU Individual indirect effect | 0.152 | 0.039 | 0.495 | 0.028 | 0.432 |
| AIL→PC→IU Individual indirect effect | 0.103 | 0.048 | 0.224 | 0.024 | 0.174 |
AIL AI Literacy, PEOU Perceived Ease of Use, PU Perceived Usefulness, PC Perceived Credibility, IU Intention to Use
Path Coefficient: The size of the effect. Bigger = stronger relationship.
Bias-Corrected Confidence Interval (Lower/Upper): If zero is not in the range, the effect is statistically significant.
Percentile Confidence Interval (Lower/Upper): Another way to double-check if the effect is real. Same logic: if 0 is not in the interval, it’s significant.
This table tells us whether things like trust, usefulness, and ease of use help turn AI literacy into actual intention to use AI. And it gives us proof that those bridges are real as long as zero is not inside the interval.