Page MenuHomePhabricator

games using Wikidata's data
Open, MediumPublic

Description

Games are an engaging way to get to know Wikidata. We should have more of them.

Things to keep in mind:

  • We do not want gamification!
  • We do not want to incentivice people to do bad edits and other actions on Wikidata

Some great existing games:

Inspirations for new games:

  • a memory game
  • http://kevan.org/wikitext
  • bring a set of items into the right order (for example order the given events chronologically or actors by number of movies)
  • https://cities.k-nut.eu
  • 20 questions
  • guess the artist (give a list of artist names and gradually reveal a painting. let the player select the right artist)
  • catch phrase (give a catchphrase and let the player guess who it is from)
  • named after (give the player a famous person and images of things. let them guess which of the things is named after the person.)

Steps to get started on a new game:

Related Objects

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptMay 12 2017, 3:58 PM
abian added a subscriber: abian.May 13 2017, 9:41 AM

To avoid bad edits with games, it would be interesting to combine the following strategies:

  • Reject statements that violate a property constraint.
  • From time to time, repeat questions that a certain user has previously answered. If, most of the time, a certain user provides different answers to identical questions, that user should be prevented from playing the game.
  • From time to time, ask questions with known answers (those already available in Wikidata, using a reference and not violating any property constraint). If a certain user usually provides answers that don't match to these, the user should be prevented from playing the game.
  • From time to time, include stupid, fake options as possible answers between the actual ones. If a certain user often selects the stupid answers, that user should be prevented from playing the game.
  • Show all the possible answers in random order each time.
  • Always include a visible option "I don't know", "Skip", "Ignore question", etc.
  • Use redundancy between different users and define an absolute minimum number of identical answers and a minimum percentage of identical answers divided by the total number of answers for each problem that users have to solve. Having exceeded these two minimums, the corresponding answer should be included in Wikidata as an actual edit. This strategy can be useful in the following situations:
    • The user interface lets users answer many questions in a short period of time.
    • The game isn't using other mechanisms to ensure data quality.
    • There's a very limited number of possible answers (e.g. male or female, yes or no, etc.).
    • There's a limited number of different questions to be shown.
  • Give detailed textual explanations each time that a certain user makes a bad decision. This have two desired effects: guides and helps good-faith users, who are interested in collaborating in a constructive way; and makes vandals angry, as they just want to have fun without caring about data quality. To increase this effect, prevent users from closing some warnings for a few seconds, time during which good-faith users will be reading the text.

Game suggestion: perhaps something like Anno Domini? Bring a set of events into the right chronological order. (Events would be either items with their own “point in time”, or the “date of birth”/“date of death”/“start time”/“inception”/… of other items.)

srishakatux updated the task description. (Show Details)May 3 2018, 1:32 AM
Restricted Application removed a subscriber: Liuxinyu970226. · View Herald TranscriptMay 18 2018, 5:58 PM
JJMC89 reopened this task as Open.May 18 2018, 6:11 PM
TJones added a subscriber: TJones.May 19 2018, 6:01 PM

Game suggestion (to show people how much Wikidata knows, rather than to generate edits or new data): Twenty Questions, which is a guessing game where one player thinks of a thing and the guesser tries to figure out what it is with up to 20 yes/no questions. It's been turned into a handheld game called 20Q which learned what questions to ask by having people play on a website and asking for new questions when it lost the game. It seems like the hierarchy of Wikidata properties would provide a similar way to find a specific thing based on careful guessing—and guesses wouldn't have to be binary, either.

abian added a comment.May 19 2018, 7:16 PM

... Twenty Questions, which is a guessing game where one player thinks of a thing and the guesser tries to figure out what it is with up to 20 yes/no questions. ... It seems like the hierarchy of Wikidata properties would provide a similar way to find a specific thing based on careful guessing—and guesses wouldn't have to be binary, either.

I love the idea! Although I guess that designing an effective, general 20Q on Wikidata will be a challenge. Wikidata has around 50 million items and 20 yes/no questions only let us distinguish 2^20 = 1048576 things, so we would probably have to select the most relevant items, 1% or less, in advance, and/or make the most of non-binary guesses, as you suggest. Also, the number of items having all the data to answer a certain list of 20 relevant questions may not be enough. Sadly, people often consider classes (the idea of "chair", "table", "computer", "book", "goat", etc.), while Wikidata classes have almost no useful statements for this game apart from the "subclass of" (P279) hierarchy.

TJones removed a subscriber: TJones.Feb 26 2019, 3:27 PM