Data tales – part five: ‘not so smart’ computing?
Guest blogger Antarin Chakrabarty continues his data tales by questioning whether all the cool, flashy and often unnecessary ways we use our computers have distracted us from their primary function – to take over the boring, repetitive tasks and give us the freedom to focus on more meaningful and creative pursuits.
People seem to think computers can magically solve the complex challenges facing our cities. No, they cannot. In fact, the computer – even the mightiest super computer – cannot do anything at all. It is simply a tool waiting for a command.
So much for the buzzwords and eye candy loving fanboys of Smart Cities in India, hoping that their gadgets will fix their cities. And yet, it is these very worshippers of digital technologies who treat the computer more as a magician’s wand than a tool of science. And who can botch up what is essential for the tool to perform its task.
Nowhere does this happen more than in the mad craze for collecting digital data. Some practitioners forget that GIGO doesn’t stand for ‘garbage-in-gold-out, but ‘garbage-in-garbage-out’.
Rampant violations of data collection guidelines, unfamiliarity with the limits of computing, ignorance of open-source software, lack of clarity regarding project goals and, last but not least, the pathological culture of sycophancy and rigid hierarchy (which makes it impossible to create feedback-based learning and correction) – all play their role in creating the illusion that smart digital solutions are at play. When in reality a relentless stream of digital garbage is being produced.
When the garbage dumped by these digital systems becomes evident as such, governments simply keep the websites and dashboards running for show, while the actual work reverts to manual mode.
So instead of reducing repetitive and exhausting manual labour, the use of such ‘smart’ technologies secretly increases it manifold. Using computers to maintain a superficial appearance of modern technology, while breaking the backs of municipal workers with manual work is some perverted application of ‘smartness’.
It is rocket science – and it’s in your pocket
I’m not much of a fan of TED Talks. Yet, there was something that the speaker said in this clip which was impossible to ignore. He showed a photo of the Apollo Guidance Computer, created by NASA engineers to do the complex calculations necessary for the Apollo 11 moon-landing mission.
He then flipped out a smart phone from his pocket, and said that a single such phone had the processing power to do the calculations for one million moon-landing missions simultaneously! He went on to show a stupid little video game on the same phone to demonstrate how we make use of this powerful tool…
The triumph of idiocy
Let me give an example of how Titusz Bugya, my Hungarian earth-scientist colleague (who taught me practically all I know about computers), and I tried to use computing to ease the burden of manual tasks in the implementation of Jaga Mission – the state-wide, slum upgrading programme in Odisha, India.
The thing about informal settlements is that they are well... informal. We have little idea what kind of land parcels will emerge. Slum dwellers don’t study cadastral records while building their homes.
However, there are administrative limits to the ability to grant land-rights to slum households, depending on the type and ownership of the land parcels on which they are located. We were aware that a number of slum houses surveyed may fall on forest land types, which would not be possible to settle without the collaboration and approval of the Indian Forest Service.
But how to identify those houses? Jaga Mission covers 168,000 households in 1,725 slums in 109 cities! Thankfully, the mission had created a digital database of all these slum houses, and also acquired digital versions of cadastral maps from the revenue department.
Ideally, all that had to be done was to create a filter in the cadastral dataset for forest category parcels and then let the computer calculate which slum houses lay within those parcels.
However, the contract period of the private technology companies was already over and the government staff (including contracted consultants) that worked for Jaga Mission had neither the knowledge to perform these tasks, nor possessed the open-source software with which to do them (the latter was a function of the former anyway).
All the copies of the 1,725 slum maps and their corresponding cadastral maps were, therefore, printed out, turning a splendid digital dataset into a massive pile of thousands of paper maps. Teams of retired revenue staff were engaged to assist the municipal teams of these 109 cities to physically visit the slums with copies of paper maps and check which houses lay on forest land parcels.
What a smart methodology! Whoever imagined that digital data could end up being measured in kilograms?
Due to our familiarity with open-source software, Titusz and I had no trouble in realising that a far simpler alternative existed, where the computer would do all this mundane work.
However, due to the GIGO style of digital work in India, the digitisation of maps was done by an army of semi-skilled technical professionals who, in the absence of standardised data entry guidelines, made so many typing errors when spelling forest type categories that it was impossible to filter the maps using these categories. This showed that even when huge quantities of digital data exist, the manner in which they are created precludes any possibility of useful computing.
Titusz then came up with a creative idea of writing small programmes to parse the whole dataset and create a dictionary of spelling errors of forest type categories. As it was impossible to clean up such a huge quantity of data garbage, we prepared a list of the ‘garbage categories’ instead (like using a magnet to find the needle in the haystack).
We discovered 53 spelling errors in the dataset of 1,725 slums. So instead of telling the computer to filter the cadastral data based on forest categories, we asked it to loop through this list every time a slum map was checked. In this way we could identify all slum houses that would correspond to a land parcel that had any of these spelling errors as its category type.
It took about eight minutes for the final programme to run and find the list of such houses with their exact geographic coordinates. Contrary to this, the manual process usually took about an hour to cover just one slum (provided it was a reasonably small-sized and less-dense slum).
Just think 1,725 hours versus eight minutes... a million moon-landings simultaneously.
Which method do you think the government finally opted for? The one involving paper maps of course.
Previous data tales
We invited Antarin Chakrabarty to contribute to this blog series and reflect on his experience working in Jaga Mission, the state of Odisha’s ambitious slum upgrading programme. His particular interest is how well this was served – or not served – by GIS.
- Data tales part one looks at the generation, ownership and use of GIS data to underpin the slum upgrading programme – or often the non-use
- Data tales part two reports on what the author found when he went looking for relevant data. Despite being told by government officials that no such data existed, he discovered vast resources that were not being used
- Data tales part three reveals the (unexpected) technological prowess of the urban poor
- Data tales part four shows how the lack of relevant or accurate data on climate risk and resilience in slums could be resolved by making community-based data collection the norm.