The Things You Know And The Ones You Don’t: Where Is Your Data Going?
This edition covers spying robot vacuum cleaners, touchscreens vs buttons in car central consoles, farmers' right to repair agriculture machinery and carbon offsets use(lessness)
How would you feel if you found a photo of yourself on the toilet posted on social media? Or maybe one of your children? This MIT technology review story describes how this happened to a number of people. The photos captured by an iRobot Roomba robot vacuum cleaner and shared with companies for labelling (video data are labelled – often by humans – to help train the algorithms used by the Roomba) somehow made it on a Facebook group. iRobot claims this should have never happened, but reality shows a different side to this story. Although apparently an isolated incident amongst consenting participants to a trial –how many times have you clicked “I accept the terms and conditions” without reading them - this story has some broader reaching implications, both in terms of data leakage and in terms of the increasingly invasive technology deployed in everyday items.
When iRobot announced its Roomba internet connected robot vacuum cleaner, the company’s sales pitch was all about convenience and effectiveness. The internet connected Roomba can be controlled using a smartphone app and “will clean efficiently […] while also taking advantage of the robot’s suite of sensors to adapt its pattern when necessary, seamlessly navigating under furniture and around clutter". All the usual stuff. Connecting the robot vacuum cleaner to the internet presents other advantages, one of which being that the data stream collected by the robot can be sent elsewhere. While iRobot claims this isn’t happening outside specifically designated trials, the capability to do so is there. What stands between photos of your house contents being sent to a third party is a “True/False” statement.
When expressions like Artificial Intelligence (AI), Machine Learning (ML), Cloud Computing are used, they seem to suggest limited human interaction with the data. Data are captured automatically, processed, and used in algorithms seamlessly and without intervention. This is often not the case. As A/Prof Beane at University of California, Santa Barbara puts it in the MIT review article “There’s always a group of humans sitting somewhere—usually in a windowless room, just doing a bunch of point-and-click: ‘Yes, that is an object or not an object,’”. Having a person taking photo of your belongings feels somewhat invasive. This sort of intimate details could be collected at some cost, until not too long ago, by surveillance teams of national agencies. With the advent of internet-connected gadgets, this capability became available to any private company which marketed a product offering convenience.
The Data Supply Chain
More broadly, an important concept is developing: the data supply chain. You, me, consciously or ignorantly are consenting to data collection of our behaviours and surroundings by internet connected devices, often for the purposes of improving algorithms which will (supposedly) make our lives easier. The data supply chain seems uniquely circular: customers (or environments) data capture feed data processing activities, the results of which are used to train algorithms to deliver a better product or service tailored to your needs. This simplified representation overlooks some interesting data supply chain issues.
The data supply chain can leak data, and there are few ways of addressing this at scale. Unlike traditional supply chains where products move, the data supply chain exclusively uses … well data. Data can be reproduced at (near) zero cost, and it is often challenging to track data generated or reproduced. Companies like iRobot contract dozens of companies to label their data. They may also sell some data to other entities. The more data can be sold, the better its return on investment, but how are these data used and by whom? We generally have extremely low visibility on what happens with the data once it has been captured or labelled.
A second issue is what Prof. Shoshana Zuboff at Harvard Business School calls behavioural surplus. Video, audio, location data contain more potential than just training algorithms. These data reveal information about ourselves – behavioural surplus – our patterns, lifestyle, recent purchases. Behavioural surplus is then fabricated into “‘prediction products’ that anticipate what you will do now, sooner and later”. Just because we provided the data for free doesn’t make it worthless. In fact, the prediction products emerging from these data are incredibly valuable especially for targeting ads. This may explain why Amazon bought iRobot for US$ 1.7 billion mid last year.
Think Twice
Could it be that Amazon is in fact interested in behavioural surplus from the data collected by iRobot? Why else would a retail platform interested in a company selling robot vacuum cleaners? By all accounts, it doesn’t seem to be their profitability (gross profit was at US$ 550 million in 2022 but net income was at US$ 30 million). Could it be that Internet-connected robot vacuum cleaners provide continuous, often unrestricted access into Amazon’s potential customers’ homes? Much like Alexa purports to help people out as a digital assistant, but in reality ends up suggesting Amazon products based on conversations, iRobot may provide a pathway for the retail giant to target products advertising based on what you have or don’t have in your house. Google managed advertising based on search engine entries in much the same way.
Supply chains generally seem an esoteric concept. They’re systems of people and organisations working with little awareness, visibility and perspective of one-another’s actions. The data supply chain operates in much the same way, just that we’re at its centre. The ‘we’ here is often collective or statistical but also a future ‘we’ whose direction and decisions have yet to be determined but that can be influenced. Participation in the data supply chain is still voluntary, although this is not immediately obvious. I’d encourage you to consider the pros and cons of personal participation in this data supply chain and what the potential costs are of the convenience offered by internet-connected tech.
In Other News
To Touch or To Push
A Swedish magazine studied the ease with which drivers interact with cars’ infotainment systems, testing between systems using physical controls and touchscreens. Their tests showed that drivers operating touchscreen-based infortainment systems were 40% slower at their tasks than drivers using physical controls.
Most car manufacturers have included increasingly large touchscreens at the expense of physical controls (much like it happened with smartphones). However, touchscreens require more attention than physical controls due to the missing haptic feedback. This raised some safety concerns for drivers. No doubt, we’ll need to install eye tracking technology in cars to beep at us when it’s taking too long to adjust the aircon temperature.
More technology seems to always be the answer to technology-induced problems!
Right to Repair: Ag Version
American Farm Bureau Federation (AFBF) and John Deere signed a right-to-repair memorandum of understanding which would allow farmers the “freedom to choose where equipment is repaired, or to repair it themselves, to help control costs”. Agriculture machinery vendors were probably one of the first to take internet-connected vehicles to a whole new level, capitalising on the data generated (and selling it back to the farmers themselves) but also limiting farmers ability to service their machinery through unintelligible error codes that could be read only in a dealership - much like a printer that doesn’t accept non-branded ink cartridges. It’s interesting to see this pushback and I suspect similar things will happen as adoption of internet connected cars increases.
So Many Offsets, So Little Impact
The Guardian reported on a 9-month investigation on Verra, one of the largest providers of carbon offsets, finding that “more than 90% of their rainforest offset credits – among the most commonly used by companies – are likely to be “phantom credits” and do not represent genuine carbon reductions.” Seems that the carbon offset button next to your plane ticket is just making it more expensive, not actually making the planet better.
I’m almost certain there’s a name for this type of scheme where taking money from people and companies increases the market value which attracts further interest and higher prices. I’m drawing a blank, maybe someone knows the right word?