So WHO's WINNING?

Table of Ranking

As participants' suggestions roll in, this ranking system is constantly updated every Saturday at 9pm. The ranking for the first trial however, will be updated at 9pm on Monday, Nov 14. New trials will also be available on these days!

Trial 1 Ranking

Average Human Score: 6.1/10

IG Algo (Bear5252) score: 8/10

Rank Algo.ID Pts Time
1 Horse 2347 10 52s
2 Bear 5252(IG) 8 1m55s
3 Parrot 3456 7.75 45s
4 Viper 7.5 1m10s
5 Tiger 900 7.25 2m10s
6 Penguin 2245 7 42s
7 Rat 444 6.5 34s
8 Croconosaur 2000 6 1m12s
9 Pingu 1999 4 16s
10 Wolf 555 3 45s
11 Ant 42 2 11s

Legend: Follow means if I have just/ have not / have already followed the creator of the suggested post.

Rank Algo.ID Like Cmt Follow
1 Horse 2347 Yes No Yes
2 Bear 5252(IG) Yes No Yes
3 Parrot 3456 Yes No Already
4 Viper No No Yes
5 Tiger 900 No No Already
6 Penguin 2245 No No No
7 Rat 444 No No No
8 Cronosaur 2000 No No No
9 Pingu 1999 No No No
10 Wolf 555 No No No
11 Ant 42 No No No

Trial 2 Ranking

Average Human Score: 6.84/10

IG Algo (Bear5252) score: 7.5/10

Rank Algo.ID Pts Time
1 Horse 2347 9.5 40s
2 Parrot 3456 8.5 1m20s
3 Croconosaur 2000 7.75 1m43s
4 Bear5252(IG) 7.5 1m52s
5 Penguin 2245 7 15s
6 Viper 6.5 12s
7 Rat 444 6 34s
8 Pingu 1999 5 12s
9 Ant 42 4.5 38s
Rank Algo.ID Like Cmt Follow
1 Horse2347 Yes No Yes
2 Parrot 3456 Yes No No
3 Croconosaur 2000 Yes No Yes
4 Bear 5252(IG) Yes No Yes
5 Penguin 2245 No No Yes
6 Viper No No No
7 Rat 444 No No No
8 Pingu 1999 No No No
9 Ant 42 No No No

Results Comparison

Rank Algo.ID Pt.1 Pt.2 Diff Pt.Ave
1 Horse 2347 10 9.5 -0.5 9.75
2 Parrot 3456 7.75 8.5 +1.25 8.125
3 Bear5252(IG) 8 7.5 -0.5 7.75
4 (TIE) Viper 7.5 6.5 -1 7
4 (TIE) Penguin 2245 7 7 0 7
5 Croconosaur 2000 6 7.75 +1.75 6.875
6 Rat 444 6.5 6 -0.5 6.25
7 Pingu 1999 4 5 +1 4.5
8 Ant 42 2 4.5 +2.5 3.25
N.A Tiger 900 7.25 X X N.A
N.A Wolf 555 3 X X N.A

Conclusions

IG Algorithm > Humans

The humans lost to the IG algorithm on both trials,but their score seems to be slightly improving after 1 round of feedback (from 6.1 to 6.85). However it is also to be noted that two human participants were absent in the 2nd trial (Wolf555 and Tiger 900). These two participants scored (3 and 7.25) in trial 1.

Change Strategy when Losing

Ant42, Croconosaur2000 , Rat444, Penguin2245, Viper employed another strategy by changing the topic of their suggested post in Trial 2! The average score of these participants increased from 5.8 to 6.35!

GF wins!

Although Horse2347 (my girlfriend) achieved an average score of 9.75, it is also to be noted that this participant also roleplayed the IG algorithm and thus had insight into my Explore tab. She claimed that my Explore Tab has little to no influence on her suggested posts by saying: "Bro, we've been dating for a year".

Out of the Box for Top 3

I realized that the top 3 winners in both trials either cater to my hidden interest that was not shown on the recordings or they catered to my current interests! For instance, my love for clown was only captured by Horse2347! Perhaps these posts also received points for most originality.

Acquantaince - Most Improved Award!

It is also to be noted that the most drastic score change is +2.5 by Ant42, from 2 to 4.5. Ant42 ranking remains the same for both trials nevertheless.

Recurring Topics

For both trials I was constantly suggested 4 main topics: UI/UX Web Design, Memes, Art, Animals. These 4 topics were indeed prevalent in my IG recordings of both Trial 1 and 2.

The Closer, The More Competitive

I saw a pattern of participants who are closer to me would be more likely to show their competitiveness to me. Perhaps it was their way of trying to impress me.

Error Prevention Methods

Having an assistant to ensure anonymity

Although I did not want to verify how many responses that I have received for each trial before its deadline, I had to do so in order to make sure there was enough participants. This in turn allowed me to guess which submission correspond to which participant by the time!

Having the AI suggestion be randomly chosen

The suggestions made by IG algorithm (Bear 5252), were partly influenced by my girlfriend as she chose a post from the Explore Tab. This falsely includes a human factor to the representation of AI algorithm. The AI might have been more accurate if she picked the first post on my Explore tab instead of choosing one of her liking. This is of course my fault (and not hers) as I did not take this error into consideration and instruct her properly.

Final Thoughts

Grateful for Participation

I am indeed very grateful for all participants! Although some of them DARED to refuse little tokens of appreciation in return, I managed to bribed a decent amount of them with snacks/vietnamese coffee.

Shoutout to Viper for submitting 4 suggestions

Yep this participants tried very hard to learn that I only ranked the first submission (for fairness accross the experiment). Props for their hard work that won't go unnoticed!

Wall of Suggestions

This wall will be updated once the experiment is completed. Here lays all the suggestions and their respective score and author.

Trial 1

IG Algithm: 8/10

TOP 3 POSTS

1. Horse2347, 10/10, art, clown, illustration

View this post on Instagram

A post shared by Julius Alexei (@angel.scape)

2. IG ALGORITHM - Bear 5252, 8/10, programming, css, web development

View this post on Instagram

A post shared by Aditya | Web Developer | 100k 🎯 (@developers_community_._)

3. Parrot3456 7,75, funny, meme, inglish

View this post on Instagram

A post shared by Sansss65 (@sansss65)

Trial 2

IG Algorithm,

TOP 3 POSTS

1. Horse2347, 9.5/10, Genshin Impact, art, illustration

View this post on Instagram

A post shared by jade ✨ (@say0ranarts)

2. Parrot 3456, 8.5, meme, deadline, mode

View this post on Instagram

A post shared by Digitally Baffled (@digitallybaffled)

3. Croconosaur2000, 7.75, UI, practical, design

View this post on Instagram

A post shared by UInterX • UI & UX Tips (@uinterx)