Digital Synesthesia User Studies Results

1st User Study - Proximity Sensor

First Exercise

During the first exercise the subject was to wear a blind fold and asked to identify how many objects were place on the table.

The objects were long thin rectangles placed perpendicular to the edge of the table where the subject was positioned. This way, the subject was able to concentrate only on a lateral movement.

The number of objects was randomized between 0 and 3.

This exercise was repeated 4 times. Twice with objects at 2" from the table and twice with objects at 1" from the table.

The level of sensitivity was also changed, twice at level 10 and twice at level 5. This sensitivity dial simply changes the mapping between the sensor data and the frequency range of the transducer. At 10 (low sensitivity) small differences are harder to detect but there is less noise. At 5 (high sensitivity) the small differences are more noticeable but so is the noise.

Percentage of correct Answers of 1st Exercise

Sensitivity 10 Sensitivity 5
Two Inch

0.47

0.41

One Inch

0.65

0.53

Overall

0.51

Discussion

The subjects had no experience except for using the device for 5 to 10 minutes before the start of the test.

Initially the idea was to have the user control the sensitivity level according to what they felt but this turned out to be complicated since the users had no experience with artificial sensing or with this specific sensory feedback loop. This is why I decided to tweak the test and control the sensitivity.

I expected to see a clear advantage of Sensitivity 5 over Sensitivity 10 since 5 would give a larger frequency jump for a smaller change sensed by the proximity sensor. But to my surprise, it was sensitivity 10 the one with the advantage and reported preference by subjects. Mostly due to the stability of the signal.

The results do not show an overwhelming success rate but I believe that this was a "warm up" exercise and it mostly helped the users get a better grip of the sensory experience for the following exercises.

Second Exercise

During the second exercise the subject was asked to identify the shape of an object placed in front of them.

The objects were one of three. A circle, a rectangle or a triangle. The shape was chosen at random.

This exercise was repeated 4 times. Twice with objects at sensitivity 10 and twice at sensitivity 5.

The subject was allowed to see the objects and see how they were going to be positioned on the table. This was so that the subject would not spend too much time trying to find it. This exercise required the subject to scan the space in 2 dimensions.

Percentage of correct Answers of 2nd Exercise

Sensitivity 10 Sensitivity 5
Circle

0.52

0.67

Rectangle

0.40

0.29

Triangle

0.67

0.80

Overall

0.54

Discussion

For this exercise, the user had a better sense of what the experience felt like. So there was always some confidence going in. From the first exercise, the detection of an edge was understood so the users would have in general two strategies.

Either find an edge and follow it in two dimensions and see what shape they would draw with their arm, or use a scanning lateral motion and try to compare how long the object was from edge to edge. The rectangle did worse than the others because users were quicker to say Circle or Triangle just because the sensation was new and different than expected. It was clear that many times the user would make a choice early and then any subsequent information from the sensor was "shoe-horned" to fit this mental model.

Third Exercise

During the third exercise the subject was asked to identify the shape of an object placed in front of them.

The objects were one of three. A sphere, a cube or a prism. The shape was chosen at random. (The cube object was the same as the square but at 3" from the table).

This exercise was repeated 4 times. Twice with objects at sensitivity 10 and twice at sensitivity 5.

The subject was allowed to see the objects and see how they were going to be positioned on the table. This was so that the subject would not spend too much time trying to find it. This exercise required the subject to scan the space in 3 dimensions.

Percentage of correct Answers of 3rd Exercise

Sensitivity 10 Sensitivity 5
Sphere

0.33

0.40

Cube

0.56

0.67

Prism

0.82

0.45

Overall

0.55

Discussion

This was a challenging exercise because the users had an understanding of how the sensory feedback worked and what the new objects should feel like. The truth was that the sphere and the prism had a very different signal than we thought. Both would ramp up the frequency as the users would expect but once on the apex, the sensor would overload and give a very low frequency. This explains the cube doing better than the sphere. The prism did better than all because the user would pick it with any strange new behaviour but mostly because once the subjects learned to map the very peculiar signature to the prism it was very easy to identify.

Total Percentage of Correct Answers

Percentage of correct answers across the three exercises.

Percentage of correct Answers through all exercises

Sensitivity 10 Sensitivity 5
Average

0.54

0.52

Overall

0.53

Discussion

Overall the users had a success rate of near 50%. I believe that this result is encouraging given the fact that they had no previous experience with sensory substitution devices and that the practice time with the device was very limited. I would expect that with a bit of training the results would jump significantly. Specially if the users get used to the specific signatures the objects give.

Dominant V.S. Non-Dominant Hand

The hand in which the sensing device was located was also randomized to see if there would be any advantage to using the dominant hand to control an artificial sense.

Percentage of correct Answers depending on hand used

Dominant Non-Dominant
Average

0.46

0.53

Discussion

The data shows that there is an advantage of non-dominant hand use but I'm not sure why this happens or what it means

Time taken to Guess

The subjects were given an initial 2 minutes to venture an answer if they felt confident. If not, the subject was given 2 more minutes. After a total of 4 minutes the system would be shut down and they had to give their best guess.

Percentage of correct Answers depending on time

Less than 2 min. between 2 and 4 min.
Average

0.55

0.44

Discussion

There is an advantage to the decision taken in the first 2 minutes. I found that the first "guess" tended to be correct but when the subject was not confident and would try to corroborate with more pases, they would feel confused.

Many times they reported knowing the right answer but then changing their answer when exploring for a longer time.