Results

Claim 01

Young adults are not be able to correctly recognize illustrations created using AI model and illustrations done by humans using digital software.

H

0

Young adults are able to correctly recognize illustrations created using AI model and illustrations done by humans using digital software 65% of times.

Binomial Z benchmarking test

Test

Success Rate

65% Benchmark

No Description

62%

AI + Human

49%

AI

74%

Human

p = 0.9445 H0 is rejected

False Description

p = 1.00

42%

AI + Human

37%

AI

47%

Human

True Description

p =0.8958

62%

AI + Human

68%

AI

57%

Human

Young adults are not able to correctly recognize illustrations created using AI model and illustrations done by humans using digital software 38% of times.

Claim 02

Presence of a description for an illustration influences its perceived origin.

H

0

The success rates for all 3 treatments (No description, False Description & True description) are the same.

One Way ANOVA

Test

There is a significant difference between the success rates for all 3 treatments.
(No description, False Description & True description)

False descriptions significantly affect perception of the origin of illustration, but true descriptions or no description don't have much impact.

Mean Success Rate on different description conditions

Post Hoc Analysis - Tukey test

0

5

10

No
description

Success rate

True
description

False
description

5

10

15

No
description

True
description

False
description

Success rate

p < 0.001

p < α

F = 17.4

H is rejected

0

H

0

There is no significant difference in success counts due to lack of description and the existence of a true description of the illustration

p = 0.994

p > α

Hence, failed to reject H0.

p = 0.001

p < α

Hence, H0 is rejected.

H

0

There is no significant difference in success counts due to lack of description and the existence of a false description of the illustration.

p = 0.001

p < α

Hence, H0 is rejected.

H

0

There is no significant difference in success counts due to the existence of a true description and false description of the illustration

Claim 03

Illustrations perceived as created by humans using digital software are more aesthetically appealing than those generated using AI model.

H

0

Aesthetic appeal for illustrations perceived as created using AI model and perceived as created by humans using a digital software is the same.

Paired Interval t Test

Test

Aesthetic Scores on different perceptions

0

5

10

15

Perceived
as AI

Perceived
as Human

Perceived as AI generated

Perceived as Human created

7.0

7.5

8.0

7.06

7.65

Median

Mean (95% CI)

p = 0.0015

p < α

H is rejected

0

There is a significant difference between Aesthetic appeal for illustrations perceived as created using AI model and by humans using a digital software.

Observed mean of aesthetic score Perceived as AI is 7.1 and that of
Perceived as human is 7.7 .

Hence, Claim 3 i.e. Illustrations perceived as created by humans using digital software are more aesthetically appealing than those generated using AI model stands true.

AI vs Human generated content

Quantitative Research Method

Field

Design Research Method

Duration

3 weeks

Collaborators

Hayat Tamboli |

Nachiket Nanoty

powerbrainai.com

AI vs Human generated content

Broad topic

Narrow topic

Factors affecting perceived origin of static images

Focused topic

To study whether people are able to differentiate between AI and human generated digital illustrations and the effect of description on the perception of it’s origin.

Motivation

To study whether people are able to differentiate between AI and human generated digital illustrations and the effect of description on Digital artists around the world are concerned that the evolution of AI will wipe out future generations of artists, it will make them obsolete and the majority of art will become intention less and algorithmically created.

As of now AI is prominent in the digital domain, we shall address issues with digital artwork and see whether humans are able to differentiate between AI-created and human-created artwork and which do they prefer aesthetically.

Research Questions

R01

Can an individual differentiate between a human generated digital illustration and AI generated digital illustration?

R02

Does the existence of Image description influence the perceived origin of the illustration?

R03

Does perception of origin of the illustration influence its aesthetic appeal?

Claims

Claim 01

Young adults are not be able to correctly recognize illustrations created using AI model and illustrations done by humans using digital software.

Claim 02

Presence of a description for an illustration influences its perceived origin.

Claim 03

Illustrations perceived as created by humans using digital software are more aesthetically appealing than those generated using AI model.

Methodology

Independent Variables

Source of illustration
(Binomial)

AI generated

Created by human

Description
(Nominal)

No description

False description

True description

Dependent Variables

Perceived Origin
(Binomial) (Within Subject)

AI generated

Created by human

Aesthetic Appeal

Confidence rating
for perceived origin

Previous knowledge of AI images

Confounding

Variables

Illustration by AI model

Illustration by humans

Controlled

Variables

Gender

Device

Random

Variables

Perceived Origin

Aesthetic Appeal

Confidence on
perception

Conceptual

Variables

Between Subject Design

Convenience Sampling

Young Adults

No description

Form 01

Form 02

Analysis

Form 03

False description

True description

Random selection

24

T1

T2

T3

23

20

Example Question

Future Scope

While 62% accuracy may seem reasonable, it also implies that humans are incorrect about 38% of the time.

Future research should investigate various factors that could contribute to misclassifications, including the complexity of images, the diversity of AI-generated styles, and individual differences in perceptual abilities.

Thanks!