Trusting machines is not necessarily better than trusting humans.
I have been working in the education sector for a fair amount of time now. My foray into the sector coincided with the nationwide “digitisation” spree. The government of India launched the Digital India campaign in July 2015. As with all government initiatives, the programme only really took shape in another year’s time. Then it began to trickle down to the states. I can neither confirm nor deny that the MHRD used some coercive tactics to influence the states. Lucky for this data wrangling physicist, my core skill set was suddenly in hot demand.
What this resulted in was this – state governments and associated agencies began seeking digital data from implementation partners. A typical conversation on programme planning would include statements to the tune of - “All this is fine, sir, but do you have an app? Some dashboard? Data-shayta we will get no?”
There were definite benefits to this transition. The first project that I worked on involved a data collection exercise from a handful of districts spread across the entire state. Previously, this was done using paper. Someone would then log all that data into Excel and analyse it again using Excel. When I came in I made use of an online data collection service and used Python/R to analyse the data. The benefit? What previously took 4-6 weeks now only took 8 days from beginning to end.
Supervisors are overworked with collecting data, consultants are splitting hairs analysing it, and teachers are losing their autonomy one supervisor visit at a time.
The underlying belief that’s propelling this juggernaut is simple - humans in a complex system are untrustworthy.
Let me elaborate. Education is a complex system, especially in a country as huge as ours.2 Take a state education system for example. Beginning at the state project office where the top bureaucrats and consultants sit, the system expands downwards through district offices to block and cluster centres and finally to schools. Many hundreds of thousands of personnel are involved in the system. Some untrustworthy, some hardworking, some value-neutral.
In this strictly hierarchical system, a Chief Minister’s vision would get turned into the Education Minister’s decision, which would take shape into the top bureaucrat’s policy, which then would pass through the maze of Chinese whispers and hopefully reach the students in a form that vaguely resembles the original vision. This is what is called the “cascade model”.
Take a quick break here and let’s talk about bitcoin for a minute.
Bitcoin, in case you’re unaware, is a form of a cryptocurrency. There are many like it. What’s important for us to understand is the basis of cryptocurrency, i.e., block-chain. Block-chain is a clever system of bookkeeping that relies on computer algorithms to maintain ledgers in a tamper-proof manner. Any transaction between two parties is etched into collective memory forever and nothing can change it without leaving its soiled imprint behind.
Traditional financial system is a mind boggling plexus of “fiat” currencies like dollar, yen and rupee. How it all runs without breaking down everyday is probably a mystery as old as time. But what some people smarter than me seem to agree on is that this system is riddled with inefficiencies and shady practices. Crypto-evangelists seek to replace this entirely with cryptocurrency in a bid to establish “the utopia of a trustless society” as DHH put it. Yes, Elon Musk is one of them.
The conclusion I want you to draw with me here is this - the underlying belief behind cryptocurrency is that humans in a complex system are untrustworthy.
Oh, where have we heard that before?
The cascade model of education is a canvas painted in greys. Certain inefficiencies and malpractices remain hidden from plain sight, and nothing bothers a
autocrat bureaucrat more than someone taking their system for a ride.
How do we fix it then?
Well, thank heavens for Nandan Nilekani, we fix it with data!
And that’s what we’ve been doing. We collect tons of data every single day. We use that data to find out who is that drunkard teacher who’s been skipping school for the last 3 months. Then we punish him. Make sure we make an example out of him too! The near real-time data helps us track the ground-level implementation of policies. For example, the state could’ve earmarked funds for setting up libraries in schools or distributing textbooks and uniforms to students. Making this happen is largely the responsibility of the block-level officers who can often be involved in shady practices when it comes to funds. So when the top bureaucrat in the state office looks at their MIS dashboard, they can tell exactly which blocks in the state haven’t yet distributed the textbooks. And the necessary remedial action is taken promptly.
There certainly are benefits to this scheme. Now that everyone knows the big man’s watching, they’ll behave themselves and do their job well.
Except, of course, they do not.
The status of education is still poor. The good people at the research centres of the country have been assessing children every year only to find that Sumu, like her brother Apu, can’t tell 3s from peanuts and 8s from laddoos.
Why though? Why did it turn out this way? Why can’t we have good things!!!
Simple: humans are most effective when they are trusted.
A teacher is most effective in her class when she believes in her students and wishes to equip them with the skills they’ll need to excel in their chosen future. But when you get her to conduct daily assessments of students and painstakingly log all that data in a broken app3, and then arm-twist her into honesty by forcing the local supervisor to visit her classroom on a biweekly basis and log further data points about her performance in yet another broken app, and to top it all, you get both the teacher and the supervisor to upload their selfies on their respective broken apps multiple times a day – you’ve gone so far down the wrong path, you might as well rear chickens now.
What happens in a panopti-system is this -
- The “good” personnel are overworked (because too much data) and demotivated (because too little trust).
- Some “bad” personnel are identified and neutralised.
- Other “bad” personnel escape the panopticon’s sight because they’re really good at avoiding this type of thing.
The neutrals remain neutrals. They weren’t bothering anyone anyway.4
And what does this mean for student learning? It remains where it always was! What the panopti-system has brought about is a mere polite shuffle, a transient rejigging, a … plus ça change, plus c’est la même chose.
I could probably pull some cross-references from other complex systems to illustrate this point better but I am sure you know it already. When you were a student, you felt it when your parent/teacher hovered over you to make sure you were studying. At work, you’ve probably had one or heard of a micro-manager. In a relationship, you’ve probably seen some overbearing, non-trusting partners.
Lack of trust never makes anything better. Technology that seeks to replace trust with algorithmic bookkeeping will not make anything better.
“Wake up, sheeple!” 5
What’s the way out? Fuck if I know, dude. I am still figuring it out myself.
What I can say for sure is this -
- Humans are most effective when they are trusted.
- Technology is not going to “fix” education.6
- Most work is just work for work’s sake. Don’t @ me.
P.S. I’ll go out on a limb and make this claim - data-based supervision of policy implementation is a good idea only for supervising policies that involve tangible materials and not otherwise. For example, if you want to find out whether each student in the state received their textbooks, go right ahead. Did the district officials release travel allowances? Do it. Are teachers faking attendance? Sure. But when it comes to measuring student learning or teacher effectiveness… tone it down a bit, please?
Keep up with the Pedagoguey!
Check out @pedagoguey on Instagram – that is where I post bite-sized versions of these blog posts, because sometimes reading a tweet thread is all you have the energy for.
You can read all the Pedagoguey blog posts here: pedagoguey.
I am guilty of it too. But I am improving upon it. ↩︎
I have a visualisation powered article coming out on this soon. Stay tuned! ↩︎
Developed by an L1 bidder, of course. What’s an L1 bidder? Technically, the lowest bidder. Practically, the “loved one”. ↩︎
Before you come at me with your pitchforks, I admit that there are exceptions to this. I am painting a somewhat exaggerated broad stroke picture here for illustration purposes only. Perhaps we need more data on this to reliably make claims? ↩︎
Couldn’t help it. ↩︎