MIT OpenCourseWareSession 9: Dynamics of Growth of a Professional Field
MIT OpenCourseWare
Opening concerns 0:00
The discussion starts with common modeling troubles. People raise task allocation, different demand for experts and rookies, the role of alternatives, service quality, and what happens when demand is too high. Hazhir notes that these choices can shape policy levers and that some short-term allocation can be simplified if the larger question is how the field changes over time.
Model structures 8:00
He then walks through the submitted models. One uses rookies, veterans, capacity, demand, and reputation. Others split customers or projects into potential, active, and completed states, or add recognition, experience, and reputation feedback. He also points out when authors model training inside the field, when they track leaving and returning, and when a simpler allocation rule may be enough.
Project timing 22:01
A fixed pool makes field expansion harder. New work comes in through awareness, a baseline order rate, and field attractiveness. The speaker prefers naming the 5% monthly baseline as the maximum fractional order rate, with attractiveness kept between 0 and 1. Finished demand then moves through completion, where the current time setting looks more like a typical time than a true minimum.
Entries and training 28:33
The same pattern appears in novices and experts. New people enter when the field looks busy, but the speaker warns that demand fulfillment alone can hide bad reputation or excess backlog. A better signal is delivery delay versus desired delay, adjusted by attractiveness. Training is limited both by novice count and by expert time, and the two training terms are really just one degree of freedom.
Training paths 42:02
Expert-led training is treated as a separate pathway from the training that comes from doing the work itself. The two can be combined by taking the minimum, if they overlap, or by adding them, if they are independent. The harder question is how much expert time is actually available, since that should depend on how many novices need help.
Recognition and quality 47:02
The discussion then turns to quality, reputation, and recognition. Quality is tied to the mix of novices and experts, but also to how much expert time goes into training instead of project work. Recognition is modeled with memory loss and a moving target, yet this is seen as messy because it mixes two different ideas. A cleaner split would let reputation affect whether adoption speeds up or slows down, while recognition would control how fast that change happens.
Model fixes and field growth 1:02:00
The discussion starts with the problem of how to treat total population and the share that is unaware, and it says the total should be tracked only if it means all people in the modeled states. The main fix is to keep quality tied to the field against an alternative field, not to reuse one variable for two meanings. It then turns to rookies, service gap, and experience. Indicated rookies should reflect extra people needed, not the gap against the whole current stock. Experience should come from assimilation and from work, but not be counted twice. A second model separates aware people, active clients, and project demand, with quality reputation moving aware people into clients and projects feeding completion rates.
Model clean-up 1:21:01
The discussion tightens the task allocation so experts take as much work as they can, while novices take the rest. That makes active projects set a desired completion rate, with expert output subtracted from it to leave the novice target. Learning time is then tied to how much novices are actually working, so idle novices learn at the slowest rate and busy ones at the fastest.
Reputation and feedback 1:23:02
The next pass questions how joiners enter the field. Demand and alternative fields are treated as outside anchors, but the point is to keep joining tied to the mismatch between demand and supply inside the field. Quality reputation is then built from expert and novice completion rates, yet its range may be too tightly fixed by those stocks and their time constants, so it may need one more free parameter.
Diabetes model structure 1:42:04
The discussion starts with a diabetes model built around treatment adherence, health system access, and patient control. Better adherence and better access should raise control and lower complications. Complications such as cardiovascular disease then add cost and burden. The model is being laid out as a patient flow from healthy people at risk, especially those over 30, to prediabetes, then to diabetes, with branches for controlled, uncontrolled, and complication cases. Age and setting matter too, since urban and rural patients may move through the disease differently. The speaker plans to add interventions such as increased awareness and test how they change the flow. Feedback pushes the work toward a clear stock-and-flow structure first, with fewer demands to force in feedback loops right away.
Narrowing the model 2:02:02
The discussion begins with a complaint that the growth model is still too broad. Hazhir says the core issue is how households and businesses buffer against uncertainty and risk over time, but that this can quickly become a full economy model. He presses for a clearer use: prediction, theory building, or testing interventions, and asks what should be in or out.
Secondary markets and prestige 2:09:02
Oriana then narrows her project to Supreme. She describes a brand that sells cheap shirts but gains prestige through scarcity and resale, so skaters can wear it while resellers push prices far above retail. The group links this to ticket scalping and asks why the primary market does not capture that revenue, or how the secondary market sustains the brand.
Feedback on models 2:20:31
The class first clears up how the models should treat desirability, price, quality, prestige, and brand use. Hazhir notes that seeing more people with a brand can reduce desirability for some goods, but the more important issue is who is wearing it. He points to conspicuous consumption and says the field’s problem statement is now more operational. He then invites one more model before closing and turns to Hassan’s work on projects, demand, reputation, and recognition.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
