Design Sprint: Crowdsourcing

  • Where We Started
We had a crowd-sourcing idea about how to give users the ability to have an impact on an AI portion of a video game. The part of the game we wanted to focus on was computer assistants or other non-player controlled characters to have growth with feedback from users and data mining. Ideas about having a BuddyBot AI while playing online multiplayer Battle Royal games (like PUBG) to help the users fill extra slots on their teams with a teammate that wasn’t just a worthless meat shield. With observing play styles and having user feedback in the form of surveys, updates to the BuddyBot system will advance the program to a threshold that users will accept that the BuddyBot is just like inviting your friend to play.
  • Future Ideas
Taking BuddyBot from the multiplayer experience and into single player games that have companions while adventuring will also improve user experience. With many people playing the games and crowd-sourcing information to nit pick the features, BuddyBot on games like Skyrim or Fallout will not have to worry about the companion making the game more struggling, but feel like adventuring with a Buddy.
An example of the problem with companion bots was found in the game Fallout 4 by Bethesda, demonstrating that a computer controlling a dog could be useful but ends up doing the wrong thing. Here is the link to the video of the example. (The dog does not die)



  • Evaluation:
What did you ask your audience?
A very important topic of crowd-sourcing is how a campaign applies the feedback and contribution from the crowd. After playing a match of PUBG Mobile with BuddyBot, we prompt the user for a short survey. After a couple of iterations of BuddyBot, we reached out to our contributors to get further feedback. We asked our audience a couple important topics. One of those is if they feel their feedback is adequately being applied to how BuddyBot. Other topics included, our after-match survey(too long, or could have more questions), efficiency of survey, and specifically what was good and bad about our campaign.

What were some of the results?

     Keep this geared for future improvements and evolution of the prototype. 

  • Implications for Design:
    • What kinds of problems are well-suited to crowd-sourcing?
The Problems that are well suited for crowd-sourcing are problems that can be divided up into small work that does not need intense learning or past knowledge to contribute to the solution. This can be avoided in some cases by masking the problem with a game and trick the user into doing really hard work but can be strategy based. More difficult tasks that just need more computing power, like solving algorithms or tests on the cure to cancer don’t need crowd-sourcing but might need more computers on the cloud contributing to solve the dense problems. The best well-suited problems is the busy work that may go into a bigger problem, sorting through images or organizing pages.  
    • Give advice to colleagues who are interested in crowd-sourcing.
The biggest hurdle when looking into crowd-sourcing is getting your problem divided up in the way that people joining in can give the correct answers. Once this hurdle is jumped over, the solution is just a bunch of people helping away. Our advice in finding the fractions of your problems to be solved is to find what basic components make up the problem and have those be the start.
    • Include one example of a task that you surmise would not be a great crowd-sourcing task.
A task that would be poor for crowd-sourcing would be public medical diagnosing of illnesses. At this time this task should only be done by medical professionals and not by people on the internet making a bit of money. The risk of misdiagnosing the patients are too high and could result in some very dangerous situations. Leave it to the trained professionals!

  • Our Prototype
We did a video prototype of how BuddyBot would give users the ability to crowd-source the AI with surveys filled out after each match and gameplay that is data-mined for improving BuddyBot’s AI. The survey questions are designed to be quick but also collect crucial information about BuddyBot’s performance.
This is a mockup of what the survey would look like after a match, this is also demonstrated in the Prototype video.
This is the video demonstration of BuddyBot:


Comments

Popular posts from this blog

385 Final Portfolio

Crowdsourcing Primer