Table 1
Operationalization of dual tailoring in artifact analysis.
| DIMENSION | INDICATORS | DATA SOURCE | DETECTION METHOD |
|---|---|---|---|
| Interest tailoring | Sector anchoring (institution-type references); role framing (professional role labels); tailoring markers (second-person role cues, adaptation offers) | Reading logs B-K (1/5-page sample) | Dictionary-based automated counts, normalized per 10,000 words; TF-IDF cosine similarity across logs |
| Comprehension-level tailoring | Definitional scaffolding (explicit definitions, paraphrases); stepwise structures (numbered sequences); check-for-understanding cues | Three selected logs (B, C, I) | Manual extraction of comprehension-oriented prompts; scaffolding density markers per 1,000 words in triggered vs. baseline responses |
Table 2
Student perceptions of Class Companion value and acceptability (n = 24).
| PANEL A. EFFECTIVENESS RATINGS (1 = NOT EFFECTIVE, 4 = VERY EFFECTIVE). | ||||||||
|---|---|---|---|---|---|---|---|---|
| ITEM | n | MEAN | SD | NOT EFFECTIVE n (%) | SOMEWHAT EFF. n (%) | EFFECTIVE n (%) | VERY EFFECTIVE n (%) | EFF./VERY EFF. n (%) |
| Course requirements (information) | 24 | 3.25 | 0.68 | 0 (0.0%) | 3 (12.5%) | 12 (50.0%) | 9 (37.5%) | 21 (87.5%) |
| Understanding course concepts | 24 | 3.04 | 0.69 | 0 (0.0%) | 5 (20.8%) | 13 (54.2%) | 6 (25.0%) | 19 (79.2%) |
| Completing assignments | 24 | 3.38 | 0.77 | 0 (0.0%) | 4 (16.7%) | 7 (29.2%) | 13 (54.2%) | 20 (83.3%) |
| Generating quality reading materials | 24 | 3.12 | 0.8 | 0 (0.0%) | 6 (25.0%) | 9 (37.5%) | 9 (37.5%) | 18 (75.0%) |
| Assessing knowledge and skills | 24 | 2.88 | 0.8 | 1 (4.2%) | 6 (25.0%) | 12 (50.0%) | 5 (20.8%) | 17 (70.8%) |
| PANEL B. AGREEMENT ITEMS (1 = DISAGREE, 4 = FULLY AGREE). | ||||||||
| ITEM | n | MEAN | SD | DISAGREE n (%) | SOMEWHAT DISAGR. n (%) | SOMEWHAT AGREE n (%) | FULLY AGREE n (%) | AGREE (SOMEWHAT+FULLY) n (%) |
| Learned more than without AI companion | 24 | 2.96 | 0.91 | 2 (8.3%) | 4 (16.7%) | 11 (45.8%) | 7 (29.2%) | 18 (75.0%) |
| Would take another AI-supported course | 24 | 2.92 | 0.93 | 2 (8.3%) | 5 (20.8%) | 10 (41.7%) | 7 (29.2%) | 17 (70.8%) |
| AI skills increased significantly | 24 | 3.5 | 0.83 | 1 (4.2%) | 2 (8.3%) | 5 (20.8%) | 16 (66.7%) | 21 (87.5%) |
[i] Note. Percentages are based on n = 24 respondents. Means and SDs are computed on the 1–4 scales indicated in the panel headings.
Table 3
Indicators of interest tailoring across reading logs (B–K).
| LOG | B | C | D | E | F | G | H | I | J | K |
|---|---|---|---|---|---|---|---|---|---|---|
| Pages in sample (1/5) | 84 | 75 | 65 | 88 | 108 | 106 | 95 | 93 | 103 | 86 |
| Words in sample | 16153 | 16003 | 13728 | 17088 | 22034 | 21182 | 19937 | 18725 | 21090 | 18069 |
| Tailoring markers per 10k words | 74.29 | 66.86 | 54.63 | 71.4 | 45.38 | 39.66 | 44.14 | 47 | 40.3 | 38.74 |
| Sacramento State refs per 10k | 3.71 | 6.87 | 8.74 | 8.78 | 9.53 | 6.14 | 6.02 | 5.87 | 8.06 | 4.98 |
| Community college refs per 10k | 33.43 | 16.87 | 20.4 | 26.33 | 9.98 | 17.47 | 19.06 | 16.02 | 20.86 | 15.5 |
| CSU refs per 10k | 29.1 | 20.62 | 15.3 | 14.04 | 24.05 | 17.94 | 15.55 | 18.16 | 8.53 | 10.52 |
| Avg cosine similarity to other logs | 0.5 | 0.52 | 0.53 | 0.53 | 0.54 | 0.58 | 0.51 | 0.56 | 0.61 | 0.56 |
| Min cosine similarity | 0.41 | 0.47 | 0.43 | 0.46 | 0.44 | 0.5 | 0.41 | 0.44 | 0.56 | 0.48 |
| Max cosine similarity | 0.6 | 0.58 | 0.64 | 0.65 | 0.78 | 0.69 | 0.65 | 0.78 | 0.76 | 0.76 |
[i] Note. Metrics computed from a systematic 1/5-page sample of each PDF (every 5th page). Similarity is TF-IDF cosine similarity across sampled text. Values range from 0 to 1; a value of 1.0 would indicate identical vocabulary distributions, while 0.0 would indicate no shared vocabulary. In this corpus, average similarities between 0.50 and 0.61 indicate that logs share a common instructional core but diverge substantially in sector-specific and role-specific language.