## 
##     control      big_ai   big_human    small_ai small_human 
##          49          50          51          49          51

Main Analyses

Overall Pattern

Overall, the results support that people are averse to the use of AI. Across nearly all outcomes, AI conditions were rated lower than their human counterparts.

However, big teams may receive less of a penalty from disclosing the use of AI. The big_ai condition was consistently rated higher than small_ai across all DVs, and while this difference only reached significance for effort (p = .003), the pattern was uniform across outcomes.

Most importantly, supporting our hypothesis, small human teams were rated more favorably than big human teams, with significant differences emerging for originality (p = .003) and effort (p = .024), and a consistent higher mean across all other DVs. This suggests that people perceive small human teams as more personally invested and effortful than larger ones.

Small_human did not significantly differ from control on any DV (all ps > .15). Though small_human scores higher in buy, original, effective, practical and effort, means were nearly identical to control across all outcomes. Therefore, disclosing small_team information may not have a huge benefit to marketers.

## $buy
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
## -3.824 -1.306  0.320  1.320  3.694 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            4.5306     0.2561  17.689  < 2e-16 ***
## conditionbig_ai       -0.8506     0.3604  -2.360 0.019047 *  
## conditionbig_human    -0.2365     0.3586  -0.659 0.510243    
## conditionsmall_ai     -1.2245     0.3622  -3.381 0.000841 ***
## conditionsmall_human   0.2929     0.3586   0.817 0.414866    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.793 on 245 degrees of freedom
## Multiple R-squared:  0.08923,    Adjusted R-squared:  0.07436 
## F-statistic: 6.001 on 4 and 245 DF,  p-value: 0.0001271
## 
## 
## $innovat
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
## -4.510 -1.034  0.380  1.353  2.837 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           5.51020    0.21810  25.264  < 2e-16 ***
## conditionbig_ai      -0.89020    0.30690  -2.901  0.00406 ** 
## conditionbig_human   -0.86315    0.30541  -2.826  0.00510 ** 
## conditionsmall_ai    -1.34694    0.30845  -4.367 1.86e-05 ***
## conditionsmall_human -0.03962    0.30541  -0.130  0.89690    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.527 on 245 degrees of freedom
## Multiple R-squared:  0.1072, Adjusted R-squared:  0.09262 
## F-statistic: 7.354 on 4 and 245 DF,  p-value: 1.312e-05
## 
## 
## $novel
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.3469 -1.1164  0.4706  1.1224  3.1224 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            5.3469     0.2207  24.227  < 2e-16 ***
## conditionbig_ai       -1.0869     0.3105  -3.500 0.000552 ***
## conditionbig_human    -0.8175     0.3090  -2.645 0.008687 ** 
## conditionsmall_ai     -1.4694     0.3121  -4.708 4.19e-06 ***
## conditionsmall_human  -0.1509     0.3090  -0.488 0.625873    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.545 on 245 degrees of freedom
## Multiple R-squared:  0.1158, Adjusted R-squared:  0.1014 
## F-statistic: 8.022 on 4 and 245 DF,  p-value: 4.301e-06
## 
## 
## $original
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.2245 -1.1800  0.0816  1.0816  3.0816 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            5.2245     0.2257  23.150  < 2e-16 ***
## conditionbig_ai       -1.0445     0.3176  -3.289  0.00115 ** 
## conditionbig_human    -0.9696     0.3160  -3.068  0.00240 ** 
## conditionsmall_ai     -1.3061     0.3192  -4.092  5.8e-05 ***
## conditionsmall_human   0.1677     0.3160   0.531  0.59620    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.58 on 245 degrees of freedom
## Multiple R-squared:  0.1268, Adjusted R-squared:  0.1125 
## F-statistic: 8.892 on 4 and 245 DF,  p-value: 1.014e-06
## 
## 
## $effective
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.0000 -0.6939  0.3061  1.0000  2.3061 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           5.61224    0.18879  29.727  < 2e-16 ***
## conditionbig_ai      -0.61224    0.26565  -2.305 0.022021 *  
## conditionbig_human   -0.57303    0.26436  -2.168 0.031152 *  
## conditionsmall_ai    -0.91837    0.26699  -3.440 0.000684 ***
## conditionsmall_human  0.01521    0.26436   0.058 0.954178    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.322 on 245 degrees of freedom
## Multiple R-squared:  0.07263,    Adjusted R-squared:  0.05749 
## F-statistic: 4.797 on 4 and 245 DF,  p-value: 0.0009636
## 
## 
## $practical
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.1633 -0.6667 -0.1633  1.3333  2.6939 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            5.1633     0.2100  24.587  < 2e-16 ***
## conditionbig_ai       -0.5633     0.2955  -1.906  0.05780 .  
## conditionbig_human    -0.4966     0.2941  -1.689  0.09253 .  
## conditionsmall_ai     -0.8571     0.2970  -2.886  0.00425 ** 
## conditionsmall_human   0.1505     0.2941   0.512  0.60934    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.47 on 245 degrees of freedom
## Multiple R-squared:  0.06159,    Adjusted R-squared:  0.04627 
## F-statistic:  4.02 on 4 and 245 DF,  p-value: 0.003545
## 
## 
## $useful
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.1837 -1.0980 -0.0408  0.9592  2.9592 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           5.18367    0.22018  23.543  < 2e-16 ***
## conditionbig_ai      -0.84367    0.30982  -2.723 0.006931 ** 
## conditionbig_human   -0.49740    0.30831  -1.613 0.107964    
## conditionsmall_ai    -1.14286    0.31138  -3.670 0.000297 ***
## conditionsmall_human -0.08563    0.30831  -0.278 0.781435    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.541 on 245 degrees of freedom
## Multiple R-squared:  0.07491,    Adjusted R-squared:  0.05981 
## F-statistic:  4.96 on 4 and 245 DF,  p-value: 0.0007327
## 
## 
## $effort
## 
## Call:
## lm(formula = as.formula(paste(dv, "~ condition")), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -3.3673 -0.8824  0.1176  1.1176  2.6327 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            5.4898     0.1972  27.833  < 2e-16 ***
## conditionbig_ai       -0.6098     0.2775  -2.197   0.0289 *  
## conditionbig_human    -0.4310     0.2762  -1.560   0.1200    
## conditionsmall_ai     -1.1224     0.2789  -4.024 7.63e-05 ***
## conditionsmall_human   0.3926     0.2762   1.421   0.1565    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.381 on 245 degrees of freedom
## Multiple R-squared:  0.1256, Adjusted R-squared:  0.1113 
## F-statistic: 8.796 on 4 and 245 DF,  p-value: 1.19e-06

Plot Means by Condition

Pairwise Comparisons

##           dv                contrast   estimate        SE   t.ratio
## 10       buy  small_ai - small_human -1.5174070 0.3586386 -4.231020
## 3        buy      control - small_ai  1.2244898 0.3622072  3.380634
## 7        buy    big_ai - small_human -1.1435294 0.3568048 -3.204916
## 8        buy    big_human - small_ai  0.9879952 0.3586386  2.754849
## 50 effective  small_ai - small_human -0.9335734 0.2643607 -3.531439
## 43 effective      control - small_ai  0.9183673 0.2669912  3.439692
## 80    effort  small_ai - small_human -1.5150060 0.2761890 -5.485396
## 73    effort      control - small_ai  1.1224490 0.2789372  4.024020
## 77    effort    big_ai - small_human -1.0023529 0.2747768 -3.647880
## 79    effort big_human - small_human -0.8235294 0.2734131 -3.012033
## 13   innovat      control - small_ai  1.3469388 0.3084461  4.366853
## 20   innovat  small_ai - small_human -1.3073229 0.3054071 -4.280591
## 11   innovat        control - big_ai  0.8902041 0.3069000  2.900633
## 12   innovat     control - big_human  0.8631453 0.3054071  2.826212
## 17   innovat    big_ai - small_human -0.8505882 0.3038456 -2.799410
## 23     novel      control - small_ai  1.4693878 0.3121135  4.707864
## 30     novel  small_ai - small_human -1.3185274 0.3090384 -4.266549
## 21     novel        control - big_ai  1.0869388 0.3105490  3.500056
## 27     novel    big_ai - small_human -0.9360784 0.3074582 -3.044571
## 40  original  small_ai - small_human -1.4737895 0.3160139 -4.663685
## 33  original      control - small_ai  1.3061224 0.3191584  4.092395
## 37  original    big_ai - small_human -1.2121569 0.3143981 -3.855484
## 39  original big_human - small_human -1.1372549 0.3128378 -3.635286
## 31  original        control - big_ai  1.0444898 0.3175586  3.289124
## 32  original     control - big_human  0.9695878 0.3160139  3.068181
## 60 practical  small_ai - small_human -1.0076030 0.2940575 -3.426551
## 53 practical      control - small_ai  0.8571429 0.2969835  2.886163
## 63    useful      control - small_ai  1.1428571 0.3113768  3.670335
## 70    useful  small_ai - small_human -1.0572229 0.3083090 -3.429102
##         p.value sig
## 10 3.156746e-04 ***
## 3  7.448889e-03  **
## 7  1.316888e-02   *
## 8  4.911614e-02   *
## 50 4.460617e-03  **
## 43 6.109395e-03  **
## 80 1.019615e-06 ***
## 73 7.225348e-04 ***
## 77 2.958697e-03  **
## 79 2.375542e-02   *
## 13 1.798297e-04 ***
## 20 2.575234e-04 ***
## 11 3.282105e-02   *
## 12 4.043571e-02   *
## 17 4.352781e-02   *
## 23 4.103110e-05 ***
## 30 2.728685e-04 ***
## 21 4.971751e-03  **
## 27 2.156244e-02   *
## 40 4.994330e-05 ***
## 33 5.518319e-04 ***
## 37 1.380087e-03  **
## 39 3.094905e-03  **
## 31 1.006009e-02   *
## 32 2.008531e-02   *
## 60 6.386748e-03  **
## 53 3.419568e-02   *
## 63 2.729570e-03  **
## 70 6.332025e-03  **

Interpretation

Overview

A consistent pattern emerged: the small_ai condition tended to produce the most negative evaluations, while small_human and control were generally the most positive.

Purchase Intention (buy)

Participants in the small_ai condition reported the lowest purchase intention. In contrast, the small_human condition did not differ from control, suggesting that small team size only penalized purchase intention when the team was AI-composed.

Significant pairwise comparisons:

  • small_ai < control (p = .007)
  • small_ai < small_human (p < .001)
  • small_ai < big_human (p = .049)
  • big_ai < small_human (p = .013)

Innovativeness (innovat)

The small_ai condition received the lowest innovativeness ratings. Notably, both big team conditions were also rated lower than control, but the small_human condition was not, suggesting AI attribution — especially in small teams — drove the innovativeness penalty.

Significant pairwise comparisons:

  • small_ai < control (p < .001)
  • small_ai < small_human (p < .001)
  • big_ai < control (p = .033)
  • big_human < control (p = .040)
  • big_ai < small_human (p = .044)

Novelty (novel)

The small_ai condition was rated as the least novel. Both AI conditions were penalized relative to control and small_human, while the small_human condition did not differ from control.

Significant pairwise comparisons:

  • small_ai < control (p < .001)
  • small_ai < small_human (p < .001)
  • big_ai < control (p = .005)
  • big_ai < small_human (p = .022)

Originality (original)

The small_ai condition received the lowest originality ratings. All AI and big-team conditions were penalized relative to control and small_human, while small_human was rated on par with control.

Significant pairwise comparisons:

  • small_ai < control (p < .001)
  • small_ai < small_human (p < .001)
  • big_ai < control (p = .010)
  • big_ai < small_human (p = .001)
  • big_human < control (p = .020)
  • big_human < small_human (p = .003)

Effectiveness (effective)

The small_ai condition received the lowest effectiveness ratings. Small_human did not differ from control, again indicating that AI attribution in small teams specifically drove the penalty.

Significant pairwise comparisons:

  • small_ai < control (p = .006)
  • small_ai < small_human (p = .004)

Practicality (practical)

The small_ai condition was rated as the least practical, significantly below control and small_human. No other pairwise comparisons reached significance.

Significant pairwise comparisons:

  • small_ai < control (p = .034)
  • small_ai < small_human (p = .006)

Usefulness (useful)

The small_ai condition received the lowest usefulness ratings. Big_ai was also rated lower than control, whereas small_human did not differ from control.

Significant pairwise comparisons:

  • small_ai < control (p = .003)
  • small_ai < small_human (p = .006)
  • big_ai < control (p = .007)

Effort (effort)

The small_ai condition was perceived as requiring the least effort. This suggests that participants inferred that small AI teams expend less effort, even when evaluating the same product.

Significant pairwise comparisons:

  • small_ai < control (p < .001)
  • small_ai < small_human (p < .001)
  • small_ai < big_ai (p = .003)
  • big_human < small_human (p = .024)

AI Attribution (ai)

Both AI conditions received significantly higher AI attribution scores than their human counterparts, confirming that participants successfully differentiated AI from human conditions.

Significant pairwise comparisons:

  • small_ai > small_human (p < .001)
  • big_ai > small_human (p < .001)
  • big_ai > big_human (p < .001)
  • big_human < small_ai (p < .001)
  • control < small_ai (p = .039)

Summary

Small Human vs. Big Human

Across all DVs, small_human consistently scored higher than big_human. However, this difference only reached significance for:

  • Originality: small_human (M = 5.39) vs. big_human (M = 4.25), p = .003
  • Effort: small_human (M = 5.88) vs. big_human (M = 5.06), p = .024

For all other DVs the pattern was in the same direction but non-significant:

DV small_human M big_human M
buy 4.82 4.29
innovat 5.47 4.65
novel 5.20 4.53
effective 5.63 5.04
practical 5.31 4.67
useful 5.10 4.69

Small Human vs. Control

Small_human did not significantly differ from control on any DV (all ps > .15). Means were nearly identical to control across all outcomes:

DV control M small_human M
buy 4.53 4.82
innovat 5.51 5.47
novel 5.35 5.20
original 5.22 5.39
effective 5.61 5.63
practical 5.16 5.31
useful 5.18 5.10
effort 5.49 5.88

But the small_human scores higher in buy, original, effective, practical and effort.

Big AI vs. Small AI

Across all DVs, big_ai consistently scored higher than small_ai. However, this difference only reached significance for:

  • Effort: big_ai (M = 4.88) vs. small_ai (M = 4.37), p = .003

For all other DVs the gap was numerically present but non-significant:

DV big_ai M small_ai M
buy 3.68 3.31
innovat 4.62 4.16
novel 4.26 3.88
original 4.18 3.92
effective 5.00 4.69
practical 4.60 4.31
useful 4.34 4.04

Manipulation Check

How large do you think the team is? (1-7) The manipulation check was successful. Participants in the big team conditions perceived a significantly larger team size compared to those in the small team conditions (both ps < .001). There was no significant difference between the two big conditions (big_ai vs. big_human: p = .533), nor between the two small conditions (small_ai vs. small_human: p = .806)

## 
## Call:
## lm(formula = check ~ condition, data = data_no_control)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
## -3.510 -1.120 -0.120  0.880  3.608 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            4.1200     0.2053  20.065  < 2e-16 ***
## conditionbig_human     0.3898     0.2890   1.349    0.179    
## conditionsmall_ai     -1.4669     0.2919  -5.026 1.12e-06 ***
## conditionsmall_human  -1.7278     0.2890  -5.980 1.03e-08 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.452 on 197 degrees of freedom
## Multiple R-squared:  0.2877, Adjusted R-squared:  0.2769 
## F-statistic: 26.53 on 3 and 197 DF,  p-value: 1.863e-14

##  contrast                estimate    SE  df t.ratio p.value
##  big_ai - big_human        -0.390 0.289 197  -1.349  0.5329
##  big_ai - small_ai          1.467 0.292 197   5.026  <.0001
##  big_ai - small_human       1.728 0.289 197   5.980  <.0001
##  big_human - small_ai       1.857 0.290 197   6.393  <.0001
##  big_human - small_human    2.118 0.288 197   7.365  <.0001
##  small_ai - small_human     0.261 0.290 197   0.898  0.8057
## 
## P value adjustment: tukey method for comparing a family of 4 estimates

AI Attribution Check

Participants in the AI conditions (big_ai, small_ai) were expected to rate the product as more AI-generated than those in the human conditions (big_human, small_human).

## 
## Call:
## lm(formula = ai ~ condition, data = data_no_control)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.1000 -1.2041  0.1765  1.1765  3.4510 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            5.1000     0.2232  22.854  < 2e-16 ***
## conditionbig_human    -1.2765     0.3140  -4.065 6.94e-05 ***
## conditionsmall_ai      0.1041     0.3172   0.328    0.743    
## conditionsmall_human  -1.5510     0.3140  -4.939 1.67e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.578 on 197 degrees of freedom
## Multiple R-squared:  0.1832, Adjusted R-squared:  0.1708 
## F-statistic: 14.73 on 3 and 197 DF,  p-value: 1.078e-08

##  contrast                estimate    SE  df t.ratio p.value
##  big_ai - big_human         1.276 0.314 197   4.065  0.0004
##  big_ai - small_ai         -0.104 0.317 197  -0.328  0.9878
##  big_ai - small_human       1.551 0.314 197   4.939  <.0001
##  big_human - small_ai      -1.381 0.316 197  -4.374  0.0001
##  big_human - small_human    0.275 0.312 197   0.879  0.8160
##  small_ai - small_human     1.655 0.316 197   5.243  <.0001
## 
## P value adjustment: tukey method for comparing a family of 4 estimates

The AI attribution check was successful. Participants in the AI conditions rated the product as significantly more AI-generated than those in the human conditions (both ps < .001), confirming that participants successfully differentiated AI from human conditions.

Control

Before today, how familiar were you with E Ink display technology used in digital photo frames? (1-7)

Familiarity does not differ by condition, and the results stay the same after controlling for familiarity.

## 
## Call:
## lm(formula = familiar ~ condition, data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.8627 -1.1800 -1.1224  0.8776  4.8776 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            2.5918     0.2565  10.104   <2e-16 ***
## conditionbig_ai       -0.4118     0.3609  -1.141    0.255    
## conditionbig_human     0.2709     0.3592   0.754    0.451    
## conditionsmall_ai     -0.4694     0.3628  -1.294    0.197    
## conditionsmall_human  -0.4154     0.3592  -1.156    0.249    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.796 on 245 degrees of freedom
## Multiple R-squared:  0.02641,    Adjusted R-squared:  0.01052 
## F-statistic: 1.662 on 4 and 245 DF,  p-value: 0.1595

##  contrast                estimate    SE  df t.ratio p.value
##  control - big_ai         0.41184 0.361 245   1.141  0.7846
##  control - big_human     -0.27091 0.359 245  -0.754  0.9432
##  control - small_ai       0.46939 0.363 245   1.294  0.6952
##  control - small_human    0.41537 0.359 245   1.156  0.7761
##  big_ai - big_human      -0.68275 0.357 245  -1.911  0.3145
##  big_ai - small_ai        0.05755 0.361 245   0.159  0.9999
##  big_ai - small_human     0.00353 0.357 245   0.010  1.0000
##  big_human - small_ai     0.74030 0.359 245   2.061  0.2405
##  big_human - small_human  0.68627 0.356 245   1.930  0.3042
##  small_ai - small_human  -0.05402 0.359 245  -0.150  0.9999
## 
## P value adjustment: tukey method for comparing a family of 5 estimates