Building Together #1: Quick-play and review

Etiene Dalcol, January 11, 2021

Building together is part of our new series of blog posts based on two concepts: Building in Public and RFCs (request for comments).

Building in Public is a growing trend towards full transparency in organizations. Companies which share everything about their financials build trust and attract talent. Indie projects help each other and gather feedback. While this is also great for marketing it is also a way of demystifying business processes.

An RFC is one possible way to start a system’s design process. There are formal and more informal ways of doing it, but in essence it is a way of gathering feedback and exploring options around an idea. Research is done, main points and issues are written down, and others are invited to comment and contribute to the document in order to find the best applicable solution. I find it very useful as an asynchronous communication equivalent of what would have been a series of meetings to resolve a topic. In many ways I find it better because it allows for proper time of research and reflection and, unlike a meeting, it is self-documented.

Quick Play / Review goals

Problems to solve

Solution

Create a quick play function with activities and exercises built with sentences players have written in the past.

Example: reorder words, fill the gap, crosswords, connect word pairs, etc.

Unknowns:

Threats and challenges

Is this a privacy violation?

Not a problem for texts that were already sent to you or your own texts. But concerning other texts, what if a player really intended for something to be seen by one player only? The privacy policy already allows using past data to improve the game, but this feels a bit different than using past match data to build word suggestions per image because there the data is aggregated and anonymised. Can this threat be mitigated?

Possible mitigations:

  1. Should exercises be built only from sentences written on challenges because those are already public to other players for voting?
  2. Can we just inform players that their match sentences are used to create more exercises for other players? Will this apply for sentences written in the past?
  3. Can we just hide the player name for older sentences?
  4. Can we just send the player a question asking if we can use their past sentences?

Not all sentences are necessarily correct

Learners can correct each other, but they might not notice when something is wrong.

Possible mitigations:

  1. Make correcting part of the exercise. Ex. “Re-order words and fix any problems you find”. Benefits: it communicates to the player the sentence might not be correct, less complex to implement.

  2. Have a consensus system for best sentences to come into exercise pipeline. Ex. Exercise with " is this sentence correct or not", challenges with best notes, sentences bookmarked, another vote, sentences written by native speakers, etc. Con: Some of these sound like a complex implementation.

  3. Use an external source of text for exercises. Ex: open books. Con: doesn’t increase feedback on players' texts, needs to be done separately per language, need to hunt for material. Pro: allows bigger context sentences and different exercises, like summarizing.

  4. Combination of above, starting with the simplest one

Some questions for you:

  1. What do you think about the privacy concerns?
  2. Can you tell us another example of exercises and activities that we can we do with sentences written by other players in the past?
  3. What’s your favorite exercise / mentioned so far in the text or comments?
  4. Can you foresee any other challenge or threat?

Recommended readings: Why ‘building in public’ is a feminist act


Etiene is a Software Engineer who is passionate about languages. She recently finished her MSc. in Computer Science researching Computational Linguistics and Computer Assisted Language Learning.