Overall GBS-building cost

It is hard to accurately identify the cost of building any one GBS accurately. The overall time or person-hours to build a GBS is a deceptive measure since projects tend to have unique scopes (i.e., overall size of content and breadth of coverage) and large numbers of variables, such as:

• Team experience; content veterans tend to move quickly, manage more information easily, and require less training to use tools.

• Clarity of initial design and teaching goals; some projects start with precise scope, but most spend a few months experimenting before settling on goals.

• Access to experts, both the physical and time constraints on the efforts and their ``turnaround''; some experts are very helpful and others more aloof.

• Availability of approving body (such as the sponsor or designer); projects frequently faltered while waiting for approval to go on to the next phase of development. Also, the less often approval is sought, the more time before each meeting spent in ``demo'' mode, which can often be counterproductive work (cleaning up temporary artwork, capturing footage badly).

• Project latency, which is a neologism I use to describe the complexities of managing three to seven people working on a project; many of these workers (like artists and programmers) may be working on many different projects at the same time.

• Contingency factors; authors can spend weeks or months on technical glitches with video or art.

Instead, some projects move smoothly and others flounder for many different reasons unrelated to their tool experience. Even so, we can see that INDIE certainly makes for cheap system construction as compared to raw code or simple interface builders.

For example, Rembrandt, Nutrition, and several student projects were all built in parallel with one another over the period of about 12 months. INDIE 2.0, which had started only a few months before these programs began, had only three ``coders'' (i.e., people who programmed): two graduate students and one programmer, and none of them were working on INDIE full-time. With only this programming expertise, a large number of non-programmers were able to work with INDIE. On projects of similar scale at the Institute for the Learning Sciences, at least one full-time programmer was assigned to each GBS. Thus, we see a straight monetary savings for GBSes--fewer expensive programmers hired, while cheaper non-programmers can produce software.

 

Table 6.4: Question, answer, point, and total objects for each project.

Project	Questions	Answers	Points	Total objects
Immunology	157	274	136	1318
Rembrandt	471	689	618	3145
Nutrition	621	903	1612	5863

Another effect of INDIE tool is that as GBS designers have gotten more sophisticated, they have built increasingly complex INDIE GBSes. As seen in Table 6.4, Rembrandt has over six times the number of evidence points versus Immunology, while Nutrition has over ten times as many points, yet all three were completed in a similar amount of time with the same size staff. For other kinds of objects, such as questions, answers and overall object counts, we see similar increases, though not quite as steep. Thus, INDIE authors have managed to increase complexity and richness in interaction each time for the same cost.

It has still taken us a year to build each GBS with a small team of non-programmers. We haven't built an Immunology-sized GBS in four months, but we've been able to produce a GBS that is 3 times the size of Immunology in the same time it took us to build Immunology in INDIE 1.0. This seems to be a common problem when developing tools--timeline compression rarely takes place. A manager for a prominent consulting company told me a joke about developing large-scale software projects: ``At our company we say we can produce a baby in 1 month--we get nine women pregnant.''

Rembrandt, Immunology, and Nutrition each spent the first fourth to a third of their total construction time designing the interaction and finding the focus of the domain that they wanted to teach about. If we were to try to build a GBS in a new domain that had never been approached by INDIE before, the evidence we have gathered implies that it would take at least 3 months again to design the first scenario, even with INDIE's help. It seems, then, that the time it takes to add complexity isn't proportional to the time spent on the project.