Literacy Project/2012-05-22

From OLPC

< Literacy Project
Revision as of 22:09, 22 May 2012 by Sj (Talk | contribs)
(diff) ← Older revision | Current revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Contents

0. Recap from the field

Richard on Wonchi and Wolonchete:

R: They are farmers in Wolonchete, down the (mount). They grow potatoes, tef.
Lots of dust and rocks near where we stayed.

On typical objects in the area
R: I don't know how often the kids get into town, but we had 2 older people who came with us, with bottles for water.
  E: They are less than a mile from the main road.
R: At Wonchi: carved stools and benches
  S: We learn all sorts of things outside our experience.
M: It's good to know which pieces are new and which are familiar, and to ensure that familiar things are presented/learned
E: Let's find ways to keep drawing more data out of what's been done and learned so far.
 - Plumbing: battery discharging, clock resets, time used, use patterns
 - Basic data: videos, reports, letter learning, vocalization.
M: Two big thoughts.
 1. What are the big things we can be learning now? how to improve with new probes?
 2. (it's a surprise, until we get through 1. :-)

1. What we can learn now, new probes to try

M: basic principles: to analyze what we know, and when we can know it, about regularity of engagement.
    C: for each app, it would be great to id what is taught; what might be learned.
    M: an imperfect analysis of this would be nice. Stephanie sent a matrix...
    E: note richard's observation on app launches, below.

For specific tech challenges, see separate section below.

App launching
Some data is captuerd through funf. How accurate is it?

  • We have a snapshot view of what apps are in memory.

  Is this regular? No... though it's roughly a uniform sample. And dates reset at times.
  It starts with the one currently in the foreground. Independent of whether the app launcher was used.

  • We have data every time an app is launched (through the launcher... not via the Apps menu?)

  This is only a subset, but we have a minimum of the # of times some apps are launched. Angela has that data.

  • App switching is frequent. Kids switch all the time (and can have 15+ in memory)

  R: mostly at Wolonchete they were using the 2 new apps; went instantly to them.

  • Video scanning: kids would scan quickly to the point they wanted.

  S: we're not probing scanning now, or how much of the video was watched.
  QQ: Should we instrument/rebuild the video player? Need to set priority.

Question: can we use our data to choose a selection to focus on by popularity?
  Given these runtime snapshots: if we just take the top app on that list, aggregated; is that a reasonable proxy for popularity?
  Minus videos... this would be good to correlate to launcher data. [all? videos show up as a single app]

New probes to focus on

M: We want to figure out a way to get new probes to reinforce beginning data.
What can kids can pronounce, and do today? What can we capture conceptually?

  • We know anecdotally we can go after letters - recognition and pronunciation. That's the single most important thing we can find out and use now.

 E: we also want to go after more ambitious ideas, aiming for thinsg that may fail; to see which of them can be learned.

  • Tinkrbooks in paticular: what mroe could we do?

 Cyn: some things ahev been done 6 times, some 50 times. we could bias pobes towards things with high exposure.
   Single words, color, number, counting...
 S: Specifically: should we check if they can count unprompted, #s after 1?
 C: We could. R: older girl in Wonchi did know (11+), younger did not.
   older children may go into the village; no pretests.
 M: We should do specific visual and auditory probes.
   - probe both lowercase and uppercase letters. we should teach them something limited, but note they are exposed to lots!
   - err on the side of probing for discovering, not probing to teach
   - have audio for individual letters
   - we want colors to come first, so they will learn the task then.

  • ICS speech recognition?


Capturing great useful videos

A: We have lots of data. The videos are the best thing we can show... for impact.
MK: Why not have someone videoing twice a week at each site?
Cyn: We need both so we can show that whatever is visible in videos is scalable to 10,000 students
R: I filmed this because they say it before they touch it. (C: are they doing this now? great.)
MK: Current videos are often kids being very soft-voiced or shy; soft singing.
M: I would die for one week of complete video. then analyze the hell out of it.
E: with a somewhat sophisticated interviewer, focused on avoiding prompting.

Cyn: can we just test them using standard reading assessment tools, with a video team?
S: The danger is inadvertently doing a lot of coaching. people saying "if you want to keep using these, learn this song"
M: Are there adults coaching? MK: sometimes. R: kids launch thigns on their own to perform.
S: We should do it, but recognize that we are interfering with experimental nature a bit to help inform the next pilot.

Cyn: Let's define what we could capture, get people recording those videos as soon as possible.
R: If you want the fastest path back, tests with video are the fastest way. With an existing test.
M: Could we get one in place next week? MK: how to conduct the tests? We could get Mike, Lidet, a uni person from UAA.
MK: In my experience, people will tell you what they think you want to hear; an empiricist from UAA.
  (SJ - how about some help from a local constructivist partner? cf. Thinking Schools Ethiopia...)
S: should we send someone from the US? MK: it's a bit overwhelming. R: Mike from the city is a big hit there.

Let's define what we mean to get, and start getting data now.
A month from now we can automate the assessment in a neutral, unobtrusive way.

Data

  • Nell: I see 4 characters: babyduck, b.spider, nell, sam sleuth?
  • Matching data: everyone loves deflating balloons (color-matching)
  • Scott shows some advanced analysis on timeseries work with balloons (letter-matching) children don't always orient the same way, but manage to use the app all the same.

  M: We really want a non-timed app also... (S: hasn't worked yet)
  S: we want to set up some rgular routine (of updates?)

Specific tech challenges and ideas


Challenges
   System use
0: R notes that children have managed to do various things that should be protected by the password.
   Tracking Apps
1. App launching
2. Orientation
3. Video player
4. Camera hole
5. Funf screen probe
   Hacking startup
6. Power button
7. Filesystem corruption
8. Date reset
   System update
9. Switching to Android 4.0 "Ice Cream Sandwich"
   We could change more of the desktop; funf has to be retested.

Other ideas

  • M: Megan @ G! offered to help with data; recs wanted for how to share related work (analysis?)


  • Cyn: Speech rec apps could be very useful. (S: will look into this on ICS)


  • Cyn: can we improve internet connectivity?

  R: 3G coverage is ok out there. EVDO is faster.
  At Wonchi, on the other side of the mountain, 3G data signal was weak but voice was fine. needs the right equipment to provide a data signal. At Wolonchete, EVDO is faster. In both places, a decent antenna could improve things; could get a monthly package off the shelf for 8G/mo.

  • Can we uncover the cameras? Punch new holes in the covers for decoration, and cut out the area around the camera?

  In Wonchi they cut a chunk out of the covers to expose the cameras.
In Wolonchete they havent' done that, so its buried under the cover, and kids have tied things thru the camera hole.

  • Working with the regional education person at Wolonchete. (Mike?) He's a great guy; if we need someone to do a week's capture, organize it through him?

  A: have someone work on user-centered design through that?

  • Gathering data: can we find people to do this from OLPC? That will help avoid bureaucracy. Someone we could pay on the ground there would be perfect.


  • We are also looking for someone else to handle data analysis over the summer.

2. <second goal postponed...>


Timeline

  • New probes next week. Some regular video next week, working directly w/o uni intervention
    Local help with video translation
  • Thinking about better ways to get data back... can we get <1wk turnaround?
    Right now it takes 9 days? 1d travel + update, 1w in play, 1d gather + send.
Personal tools
  • Log in
  • Login with OpenID
About OLPC
About the laptop
About the tablet
Projects
OLPC wiki
Toolbox
In other languages