Using AI to Get simplified accurate Player Feedback

Suggestions
1

Instead of relying solely on manual surveys and subjective analyses, developers can now use advanced algorithms to sift through thousands of comments, pinpointing trends, concerns, and praises with unparalleled accuracy.
One of the greatest advantages of AI is its ability to automatically categorize feedback. Comments about graphics, gameplay, soundtrack, or bugs can be instantly classified, allowing development teams to prioritize their responses and actions. Moreover, the sentiment analysis provided by AI can assess the overall tone of comments, helping to determine whether the reception of an update or new feature is predominantly positive, negative, or neutral.

Using the latest news as a reference, I created an example of how data collection with AI can work. The example uses the language model Chat GPT-4 with the plugins: WebPilot and Wolfram.

Last post used: ✨ Jungle Fire update
Number of replies analysed: 143

RESULTS OF PLAYERS REPLIES

  1. Questions: 32 comments
  2. Positive Comments: 45 comments
  3. Negative Comments: 38 comments
  4. Suggestions: 12 comments
  5. Neutral Comments: 10 comments
  6. Off-Topic Comments: 6 comments

image

Based on the content extracted from the forum post, here are the top 10 topics discussed in the comments:

  1. Jungle Fire Update: The main topic of the forum post, many comments revolve around feedback, opinions, and discussions related to the new update.
  2. Weapons: Discussions about the new weapons introduced, their balance, and how they fit into the game.
  3. Maps: Comments about the new maps, their design, and gameplay implications.
  4. Squads: Feedback and discussions about the new squads and their impact on gameplay.
  5. Game Mechanics: General discussions about game mechanics, possibly affected by the new update.
  6. Bugs/Issues: Players reporting bugs, glitches, or issues they’ve encountered after the update.
  7. Game Balance: Concerns and discussions about the overall balance of the game after the update.
  8. Suggestions: Players offering suggestions for improvements, changes, or additions to the game.
  9. Graphics and Performance: Comments about the game’s graphics, visuals, and performance, especially in the context of the new update.
  10. General Feedback: General comments and feedback about the game, not necessarily tied to the Jungle Fire update.

After an analysis I discovered that a slight margin of error may occur for analyzes carried out using the Web Pilot extension. If people are interested in this post, I will share the conversation and create a way to analyze it in a 100% accurate way using the gpt chat code interpreter function

There are many things I can do, I need suggestions of more things to analyse.

My suggestion in this post is that developers use more efficient ways like the one I mentioned to analyze feedback and bring better results to the game.

2 Likes
2

AI is really starting to become a helpful tool. I use it to write emails, organize, and even generate and track my weekly workout routine. Saves a lot of time.

3 Likes
3

I’m kinda skeptical.

Unless the AI algorithm is able to realize when a certain topic is being purposefully spammed beyond decency by one or few subjects, whom are renowned/infamous for being extremely active on the forum, it seems quite easy to manipulate the results by spamming threads, commenting everywhere even if off topic, and exploiting buzzwords.

An AI feedback collector should be able to distinguish between actual feedback and some whiners’ endless gibberish.

1 Like
4

I understand your concerns. But this is not that difficult to avoid. I can guarantee that currently with the training that the AI ​​has, it can quite naturally filter for troll users. But I understand that it is not a 100% accurate way. Some examples of other alternatives are: Filtering by users with more than 100 days of login, feeding a manual list of troll users. The best way currently is through AI training. You just need to include some messages as an example of troll messages and the AI ​​will more effectively recognize troll users

5

I’m not referring to trolls, but rather to those who try to push their own agenda on how the game should change to fit their personal preferences, by flooding the forum with dozens of posts everyday and trying to derail every thread they step in towards their topic of choice.

1 Like
6

AI can only count one post by user to generate feedback. repeated publications are not counted

1 Like