Contribution, Attribution
Is this the same as necessary/sufficient?
Instrumental conception of information…if information is good, widen the pipe.
Another conception of
Research Pitches
Why do we care about ICTs? There’s a refrigerator divide and nobody suggested community refrigerators.
Census
Implicit assumption that these are good, but we never know how many there are.
Database of the relevant categories, types–not a survey. The survey would follow.
Need to have some kind of count. The numbers are very important.
Benchmark to use to make sense of findings.
Start with the assumption that it won’t be perfect, but it will be useful.
For private entrepreneurs, participation in the census can be self-promotion.
What would census accomplish?
Help with baseline data to measure other things against.
Measure different magnitude, categories of impact.
Geographic variables and comparisons come into play.
What would this miss?
Risks missing evolving technologies and services.
Categories could be confusing/misleading.
Would not capture what people are doing in space.
Risk of data not being equally valid across sites, countries.
Depends on the information you collect with census: must collect info on what’s happening.
Sample v. census
Can’t capture the unexpected
Obsolete results, landscape is changing fast.
3 approaches: Panel, Outcome Mapping, Most Significant Change
Public Access doesn’t come until demand is aggregated–supply side. In urban, build it and they will come. In rural, they won’t demand needs to be aggregated. Need to measure the demand.
When demanders go for supply, and activity has occurred. It has an output. 23 children learned powerpoint. Which leads to an outcome–they are empowered to give presentation. The impact comes next. The farther down this chain you go, the more difficult it is to count the impact.
Panel
Stakeholder analysis. Users, businesses, etc. go for a particular need. Attract a bunch of people (users) and track what they do as a result of access.
Outcome Mapping
Outcomes can’t be linearly linked. We are going to track outcomes only, with special emphasis on boundary partners. What bp does differently is what we’re going to track. As a result of empowerment, education etc. what do they do? Organize? Vote?
Most Sig Change
Pick a group of people and ask them in a period of time what was the most sig change. It can document emergent change.
Outcome mapping and msc both require other tools to work well. Both can work well as a structured way to organize narrative. It can also help identify indicators to be measured.
What would this miss?
Risk that this technique would be overwhelmed by context. More difficult to move to the lessons you can take away. Sampling about generalizability key. On what grounds and when do you sample. What is the procedure for sampling?
Quasi-experimental design + Retrospective
Not sure why its quasi.
10+ years of significant m&e of public access: 1) surveys of users, operators, nonusers, stakeholders, focussed on self-reporting. 2) Tech informed–what pages, how many bits, etc.
Not done a good enough job of using work of predecessors.
Experimental approach asks for a smaller step…how might we best design an intervention to best reach the outcomes we want.
ESL in schools + in community centers.
Some groups doing some system in co-located shareware vs. private environs.
2 activities, track differential down the causal chain.
Retrospective Study
Study use from programs that are 5 or 10 years from inception.
Even places that have failed, we could look at long term impacts.
Potentially cheap and easy–pick right project, do it relatively
Gets closer to attribution (from contribution) because you can control a particular factor.
Quasi because you pick something that you know will change…its experimental if you’re doing the intervention.
What do we miss?
Attribution is still tough, because many factors always in play.
Experimental design in an emergent field.
Range of design limited by factors you can control or manipulate.
Require alot of theory upfront to know what to factors to manipulate.
Ethical issues, if negative consequences.
Risk of false attribution…understand the instance but not the larger phenomenon.
Knowing what was gained by retrospective–history + cheap.



