Join 3,494 readers in helping fund MetaFilter (Hide)


InBloom wilts under privacy heat
April 22, 2014 11:57 AM   Subscribe

Controversial education tech company InBloom has shut down over student data privacy concerns. Backed with $100 million in grants from the Bill & Melinda Gates Foundation and the Carnegie Corporation of New York, InBloom quickly announced nine states (CO, DE, GA, IL, KY, LA, MA, NC, NY) as partners, with more than 2.7 million students enrolled, with the goal of using big data to direct education emphasis and other decisions. With a recent decision by New York state to halt participation in any project involving storing student data in the way InBloom had planned (and the deletion of any such data already stored), all nine states had either put data sharing plans with InBloom on hold, made them voluntary, or pulled out completely.

Parents and advocacy groups decried data collection efforts that pulled as many as 400 data points per student, including personally identifiable information such as social security numbers and intimate details about family relationships and reasons for enrollment changes. Futhur, despite assurances that "the company was designed as a non-profit from the beginning to keep special interests at bay, in order to create a tool that is data, platform and user agnostic", activists, parents and educators lead by Leonie Haimson of Class Size Matters expressed concerns that there were no guarantees that data would be secure, or that it would not be shared with third party vendors (a claim InBloom denied).

In a final message on the Inbloom website, CEO Iwan Streichenberger blamed the failure on "mischaracterization" and "misdirected criticism", even as he praised the InBlood team and the promise of data-driven education.

At a conference earlier this week, U.S. Department of Education office of technology Director Richard Culatta said that private companies need to do a better job explaining privacy policies to students and parents: "It's time to say it in plain English," Culatta said. "It's time to say it in ways that teachers and parents can understand." Streichenberger, also at the conference, remained optimistic "we've come to realize that public acceptance of what inBloom does will take a long time."
posted by 2bucksplus (29 comments total) 10 users marked this as a favorite

 
As someone running college IT systems containing FERPA-protected data, and a parent with kids whose data is in a town-wide Student Information System, I figured this would be their biggest stumbling block from the first time I heard about it.

The more data you have, the better inferences you draw -- and the wilder the privacy freak-outs by parents (myself included :7).
posted by wenestvedt at 12:00 PM on April 22 [4 favorites]


No, the problem wasn't that the privacy policies weren't explained well enough.

The problem was that there was no need to compile that level of data on students in the first place. Public acceptance of what inBloom was trying to achieve should not take a long time - it should never happen.
posted by NoxAeternum at 12:01 PM on April 22 [18 favorites]


Wow, an amazing degree of "we're sorry you're all so stupid that you won't let us bring you the huge benefits of this wonderful thing you're too dumb to know you even need," in those press releases. First I've heard of this project as I am not an education geek, but it seems pretty obvious that guys who communicate like that are going to rub some people the wrong way and have a real hard time overcoming natural wariness about their big, disruptive idea.
posted by Naberius at 12:08 PM on April 22 [9 favorites]


I don't think you two have met. Risk, meet Benefit.
posted by Mental Wimp at 12:10 PM on April 22 [4 favorites]


Over the last couple of years I've become increasingly aware of outfits peddling the notion of storing HIPAA-protected patient medical information in 'the cloud' to hospitals. That should end well.
posted by jfuller at 12:41 PM on April 22 [2 favorites]


Like with so many other fields these days, we don't need high-tech innovation in education to improve outcomes, which is absolutely nothing more than glorified PR for private firms to take public money. What we need is a commitment to properly funding models of process that we already know to be efficacious, like ensuring that everyone gets the same good education rather than segregating students into differing tracks based on their parents' socio-economic status.

We need innovation in areas where we don't know how to do better, not in areas where politics prevents the implementation of good solutions we already have.
posted by clockzero at 12:41 PM on April 22 [20 favorites]


CEO Iwan Streichenberger blamed the failure on "mischaracterization" and "misdirected criticism", even as he praised the InBlood team

Interesting slip. I wonder what his real plans were!
posted by Jpfed at 12:50 PM on April 22 [6 favorites]


Outfits peddling? It's practically mandated by law.
posted by phaedon at 12:54 PM on April 22


(The slip is mine. I typed InBlood at least 15 times while writing this post. What does that say about my plans?)
posted by 2bucksplus at 12:59 PM on April 22 [3 favorites]


"Teachers should be able to easily support the individual learning needs of students," the statement reads. "We believe the technology behind inBloom is an important part of making that a reality."

Thus the Gates Foundation. Bill Gates missed the data mining revolution so he doesn't appreciate that you basically have to be as evil as Google - hoovering up every bit of personal information you can - preferrably without your knowledge and consent - to make it work.
posted by three blind mice at 1:07 PM on April 22


> Outfits peddling? It's practically mandated by law.

Well, we (that is, the 350-bed hospital where I was PACS admin until very recently) weren't when I retired, about a year ago. We had a big fat mirrored SAN--on site--for everyday storage and business continuity and an equally big fat digital tape robot--also on site, duplicate tapes stored offsite-- for real backups. If anybody had tried to move all that to the cloudspace some company leases on that server farm in the .ru domain I would have thrown a shitfit and, if necessary, walked out. (It is actually true that I could foresee said shitfit and walkout coming down the road, though I could not be certain how near it was, and my lack of enthusiasm for being a part of the showdown was a factor in deciding to pass the PACS hat on to somebody else.)
posted by jfuller at 1:16 PM on April 22


Wow, an amazing degree of "we're sorry you're all so stupid that you won't let us bring you the huge benefits of this wonderful thing you're too dumb to know you even need,"

a friend of mine refers it to the Human Factor -- the reason why so many data-mining related tech-solutions MUST fail. Not because there's anything intrinsically wrong with the tech, but because there is something intrinsically wrong with humans. The information will be abused.

This is why we can't have perfect things.
posted by philip-random at 1:18 PM on April 22 [6 favorites]


I don't see what the big deal is. This is all data produced during educational testing. Lots of companies collect it, and they have full cooperation of schools and school boards. The volumes of data are enormous. Most of it is used by statisticians and Ed.D.s for research on the test results and inferences about the student population's educational achievement. There is not much data here of commercial worth (e.g. college preferences sold to college recruiters) that hasn't been widely available for decades.

Disclaimer: this is actually my job at the moment.
posted by charlie don't surf at 1:34 PM on April 22 [2 favorites]


sometimes the rabbit on the riverside escapes the jaws of the lunging crocodile, and the crocodile undergoes a period of dejection and hunger pangs before he finally meets with success. won't ANY of you here think of the croc?

from the "intimate details of family relationships" of student joe jones, still available after purported deletion, because potentially useful information is never deleted:

"father caught vd from a stripper."
"sister and boyfriend stole lexus, crashed it into grain silo."

it is socially beneficial for this information to be for sale by a limited number of stalwart corporations, because we need to equitably adjust your taxes and insurance premiums in our free-market driven model of transparent risk discovery. we need your social security number too, because how else could we have a "user agnostic" platform without it?
posted by bruce at 1:41 PM on April 22 [3 favorites]


I'm a big fan of more data in general for stuff like this, in large part because I often have to answer random questions from large data sets from the past, but I also acknowledge that there are privacy concerns with PII. I'd rather see that met with greater legal privacy protections in general, rather than just cutting out this educational data; I like the European idea that you own the data you produce. Stricter restrictions on the sharing of the public's data with third parties would be a good thing in general.

And I'm not sure why InBloom couldn't have continued on, since the first law passed (haven't had time to read the budget bill) just requires parental consent. It wouldn't get the full data set, but still it'd do pretty well.

(I also wonder if InBloom would have done better at pitching this as providing a way for parents to get full data without PII on their kids and classes.)
posted by klangklangston at 2:19 PM on April 22 [2 favorites]


> I don't see what the big deal is.

From what I understood about InBloom it's not that the data itself is of commercial worth but the idea of a single API fronting a whole lot of student data, intended for private sector use. What gave me pause personally was not so much the privacy concerns (all that data is already in third party SISes), but that part of the plan seemed to pretty explicitly be a streamlined access point into school budgets via lowering the bar for people to create of apps that would to crunch the data in InBloom. And while opening up record keeping for automation is in theory great and saves lots of drudgery, I recall pitches like automated IEPs assembled by tracking and predicting student performance -- cool robot utopia stuff that seems really hard to get right while not having a ton of discernible benefit over humans doing the same thing, aside from it being a service the district buys, per year, for a bit cheaper than it'd get the humans.

Which is to say I felt pretty ambivalent.

I could have sworn the original discussion I saw of it was here on MeFi (came up via something about SXSW, I think), but I can't find it.

There's developer documentation though. The data entities are fun to to look at: it's kind of neat how they have the Learning Style broken out as percentages over auditoryLearning, tactileLearning, and visualLearning, but also a reminder of how hard it can be to represent information like that either in a database or a state bureaucracy (or both). Like how the economicDisadvantaged field is just a single boolean.
posted by postcommunism at 2:29 PM on April 22 [3 favorites]


"Like with so many other fields these days, we don't need high-tech innovation in education to improve outcomes, which is absolutely nothing more than glorified PR for private firms to take public money. What we need is a commitment to properly funding models of process that we already know to be efficacious, like ensuring that everyone gets the same good education rather than segregating students into differing tracks based on their parents' socio-economic status."

...But imagine having the data to be able to actually say that, if true, and mean it with R2 and P values that would be actually convincing along with much more precise and institutionally valuable things. This could have been really cool in a lot of profoundly valuable ways, and having an ability to demonstrate what is and isn't a problem with education would have been one of the big ones. Is the success Charter Schools in some districts have had really a reflection new segregation? What predicts student success or failure after a switch move between schools? What factors really influence school violence? Do current teacher evaluation methods predict a goddamn thing? How much does the availability of specific social services in a State really affect student performance with all other factors accounted for? Which students have more medical absences during flu each season and why?

There was so much more though, imagine IEPs for everyone with diagnostically robust ways to automatically flag students in need of various kinds of attention, from academic to developmental to behavioral, who might otherwise be invisible to staff who should be helping them with demonstrably effective interventions.
posted by Blasdelb at 2:31 PM on April 22 [6 favorites]


It took some creative searching but this was the previous thread
posted by Blasdelb at 2:32 PM on April 22 [1 favorite]


Yep, that was it. (And I see that link lands directly below my fantasy of a covetous, peeping robot.)
posted by postcommunism at 2:38 PM on April 22


it's kind of neat how they have the Learning Style broken out as percentages over auditoryLearning, tactileLearning, and visualLearning

Tangent, but (sensory) learning styles have not been convincingly demonstrated.
posted by Jpfed at 2:52 PM on April 22 [3 favorites]


"Tangent, but (sensory) learning styles have not been convincingly demonstrated."
These kinds of data set could have provided a very powerful answer to that question too!
posted by Blasdelb at 2:55 PM on April 22 [3 favorites]


♫ ...a beautiful bride, a handsome groom... ♫
posted by Wolfdog at 4:24 PM on April 22 [1 favorite]


I don't have a really strong opinion on this particular case, but I can and must say I share many people's skepticism that people will properly take data quality and integrity issues into account when hoovering up and analyzing data from these massive theoretical data stores. It takes an incredible amount of care and attention to detail to get big data right, and indicators that can seem clear-cut and simple to interpret when simplified down to data points can actually mislead you about the underlying real-world situations. I wish I had time to offer some concrete examples I've seen in my own work, but it's bedtime for the kids. I'll try to drop back by and elaborate later if I get a chance, but I really think people are too inclined to make facile assumptions when interpreting aggregate data to have any great hope big data will be anything like the magic bullet a lot of folks would like it to be. That said, I don't mean to say there aren't valuable uses for the data, just I'm not sure we'll ever be capable of properly harnessing those benefits and certainly not without approaching the challenge with a hell of a lot more humility, compassion and pragmatism than we tend to bring to such things.
posted by saulgoodman at 5:03 PM on April 22 [6 favorites]


I just don't think our society values education. Exhortations to improve it almost always assume that it's merely an instrumental good, something that produces the semi-skilled workers of tomorrow. How do you change that?
posted by thelonius at 6:15 PM on April 22


...But imagine having the data to be able to actually say that, if true, and mean it with R2 and P values that would be actually convincing along with much more precise and institutionally valuable things.

You're assuming that politicians (or PTA chapters, or whoever holds power) would be convinced that they should devote resources to education more equitably if this or that could be demonstrated with R squares and P values. I think we're proceeding from different premises. My understanding of the scholarship on educational attainment is that there isn't as much mystery about how to help children learn and succeed as the questions you posed seem to imply, but maybe I am misinformed. It seems to me, also, that the central issue involved with convincing power-holders in the US about what works and what doesn't is not that they lack information but that they have to satisfy constituencies who regard spending money on Black or Latino or any-ethnicity-poor children as a waste. Is there any empirical reason to think that the lack of egalitarian educational attainment in the US is something other than an artifact of anti-egalitarian politics?

This could have been really cool in a lot of profoundly valuable ways, and having an ability to demonstrate what is and isn't a problem with education would have been one of the big ones.

Again, though, I think this contention assumes that we want for demonstrations of efficacy in the US, but it seems to be agnostic to the possibility that policy-makers don't care all that much, as a class, about what works and what doesn't for students, but instead make decisions about how education will be carried out by other principles. How could we determine which scenario is closer to reality?

Is the success Charter Schools in some districts have had really a reflection new segregation? What predicts student success or failure after a switch move between schools? What factors really influence school violence? Do current teacher evaluation methods predict a goddamn thing? How much does the availability of specific social services in a State really affect student performance with all other factors accounted for? Which students have more medical absences during flu each season and why?

There are some very good questions there, and some perhaps less pressing ones; all of them seem worth answering to me. I think it's notable, however, that none of them address the most important determinant for educational outcomes, which is simply resource allocation. Children from low socio-economic households, who aren't important enough to have much money spent on their education, have lower attainment, and that's true cross-nationally as well, but in countries with greater economic equality or an educational funding mechanism that doesn't simply reproduce extant socio-economic classes, it doesn't matter as much as it does in the US.

So while I agree that there are many questions we can still ask and answer about education, it seems a little silly to pretend that we don't already know the answers to the big and important questions, and we shouldn't mistake these other questions for them.

There was so much more though, imagine IEPs for everyone with diagnostically robust ways to automatically flag students in need of various kinds of attention, from academic to developmental to behavioral, who might otherwise be invisible to staff who should be helping them with demonstrably effective interventions.

I think all children should be able to go to schools where kind, sympathetic adults with lots of training and expertise can work with those children and their parents to help them with many different kinds of problems; that being said, I am not convinced that the scenario you describe here is only possible by means of, or ought to be accomplished by private companies collecting data at the scale inBloom intended to, and I'm especially skeptical about the implication that letting the free market do what it wants with personal data about children would both result in that scenario and avoid other, considerably more problematic ones.

I don't think the US's relatively low educational attainment is attributable to a lack of data, to summarize, and I'm not sure where that inference comes from. Technocratic interventions to education which seem to pretend that comparative resource devotion isn't worth talking about seem inherently suspect to me.
posted by clockzero at 6:20 PM on April 22 [4 favorites]


I'm not ambivalent about this issue. Collection and centralization of data like these could be an important tool in better delivering education. The same is true for medical records and medical care. And the price we should have to pay to consume the public education for ourselves or our progeny is the same price we should pay to consume the medical knowledge gained from past experience of patients: make our data available.

That said, the need to ensure that the data so collected is secured in ways that protect the individuals' data is supreme. I know this, because I spend a good deal of my professional life analyzing such medical and health data. When there is breach of privacy, it raises the level of suspicion and suddenly fewer datasets are available, the cost of collection and use of data goes up, and more bureaucracy eats up the resources allocated to research. A concrete example is access to state drivers' license and ID databases. In my state, there has been a series of scandals where police and other government workers have used these databases to look up, harass, or stalk someone for personal reasons, including pure entertainment. Because of these scandals (none involving researchers), researchers are no longer allowed access to these databases for population research purposes (e.g., surveys or recruitment to studies).

The privacy extremists, who don't seem to understand that the quality of care they receive depends upon all the prior people's experience with their or related conditions, would have you believe that each tub stands on its own bottom, and that doctors magically go to medical school and learn how to treat illness for all time without having to have access to anyone's medical records for research purposes. How this cartoon version of life came to exist in their brains, I have no idea. But it needs to be eradicated, because the commons (e.g., the educational or health experience of the population) is essential to our continued health and well-being. And those accessing these types of databases need to lose any cavalier attitudes they might have about privacy and security, because the entire enterprise rests on respect for those data.
posted by Mental Wimp at 8:45 AM on April 23 [1 favorite]


These guys are totally taking a play from the Total Information Awareness handbook. In 20 years we'll all be screaming about how invasive all of this is.
posted by Revvy at 6:28 PM on April 23


School districts are already handing over all sorts of sensitive data to smaller vendors, often with little oversight. It gets less attention because it's done piecemeal, rather than in one big data warehouse like inBloom.

It's hard to say whether the criticism of inBloom's privacy practices is legitimate, but certainly the company was, and still is, tone deaf. When parents are up in arms, you don't scoff at them and tell them they just don't get how valuable this is and it's a nonprofit and hello, GATES FOUNDATION.
posted by chickenmagazine at 6:44 PM on April 23


It's true. It's also true that a lot of work done with large and small datasets is taken very seriously and done well. But the rub is, raw data itself has to be very closely scrutinized before it can be merged, or transformed. Fields in different data sets may have different meanings but seem to refer to the same things, but definitive data dictionaries may not be available. Data sources to properly analyze a particular question may not be available, but there may be external pressures to make unjustified inferences to answers the question when the data doesn't really support it.

Not to mention, looks can just be deceiving. I remember once we had two student records, and everyone was convinced they were dupes. The names were identical, genders identical, SSNs identical; birthdate was off, but only by a few days. So we at first assumed the birthdate had been fat-fingered during data entry. Nope. Occam's Razor be damned, truth is more complicated than fiction.

Real life investigation determined the two students were cousins from an immigrant family that observed a family tradition of naming their children after recently departed ancestors. Both kids, having been born around the same time (hence, the close but not quite same birthdates), had been named after the same ancestor. They were immigrants living with an American uncle and had not yet been assigned SSNs, so both had used their uncle's SSN when signing up for the program.
posted by saulgoodman at 9:13 PM on April 23


« Older Your Friendly Neighborhood Drug Dealer...  |  "Faculty at private schools we... Newer »


This thread has been archived and is closed to new comments