Clever Updates Their Privacy Policy, and Does It Well

3 min read

Clever recently updated their privacy terms, and their method and process provides a good example for other companies to replicate.

First, they checked their privacy policy and terms of service into Github. This is a simple step that any edtech company can do. By having terms checked into git, we automatically get an annotated changelog of how terms evolve. This log is a great step toward increased transparency. All companies are using some form of version control to manage their codebase; extending this to their terms of service and privacy policy is incredibly straightforward.

In terms of the actual content, Clever made two substantial changes around how data is treated in case of a sale, transfer, or bankruptcy.

Here is Clever's original language, before the updates:

In the event of a change of control: If we sell, divest or transfer the business or a portion of our business, we may transfer information, provided that the new provider has agreed to data privacy standards no less stringent than our own. We may also transfer personal information – under the same conditions – in the course of mergers, acquisitions, bankruptcies, dissolutions, reorganizations, liquidations, similar transactions or proceedings involving all or a portion of our business.

Here is Clever's updated language, after the updates:

In the event of a change of control: If we sell, divest or transfer our business, we will not transfer personal information of our customers unless the new owner intends to maintain and provide the Service as a going concern, and provided that the new owner has agreed to data privacy standards no less stringent than our own. In such case we will provide you with notice and an opportunity to opt-out of the transfer of personally identifiable Student Data.

The updated language contains two important distinctions. First, the terms now clearly state that personal data will not be transferred as part of any transaction unless the buyer is going to maintain the service, and do so under sound policy. This goes a long way toward mitigating the risk of another ConnectEdu or Corinthian Colleges fiasco.

The second important change in this section is that the terms now guarantee notification in case of sale, paired with user opt-out of data transfers. This puts control in the hands of end users. I'm also assuming here that end users could be both schools that contract with Clever, and people who attend those schools.

The last line of the policy shows an additional minor change: the privacy policy has the date the policy was updated, and the date the updates went into effect. This is a great shift because it shows that a review window of updated terms is a standard part of doing business.

These changes get a lot of things right. By identifying that a sale to a company that wants to maintain and grow the service is a different type of event than a liquidation, Clever combines good data stewardship as a parallel to maintaining the continuity of their service. By committing to notification and user opt out, Clever demonstrates a confidence in what they offer, and a respect for the trust people place in them. By putting their terms in Github, and maintaining solid commit messages, they make it very clear how their terms evolve over time.

Most importantly, both the process that Clever has used, and the content of these changes, show how this can be done well. I'd love to see other companies follow Clever's lead here.

, ,

When Will Apple, Google, and Microsoft Require Encryption For Educational Apps?

1 min read

Microsoft, Google, and Apple all maintain app stores with sections dedicated to education.

Unfortunately, within the EdTech space, many vendors lag behind broader industry standards around security - including basic things like encrypting connections and logins via SSL.

If Microsoft, Apple, and Google all started requiring basic encryption as a condition for continued inclusion in their respective marketplaces, vendors would up their game. All marketplaces already review apps as part of their approval process. A check for basic security practice for educational apps would improve the quality of the apps, and the quality of the services offered by vendors.

This wouldn't solve every issue related to privacy and security in edtech, but it would eliminate many of the obvious weaknesses.

So, Apple, Google, and Microsoft - how about it? When will you start requiring encryption for educational app developers?

, ,

We Have An Assessment Problem

2 min read

The Civil Rights Project has an updated report on rates of suspension in schools that came out on February 23rd. The report has some great details that show the scope of the issues facing us as we attempt to dismantle the school to prison pipeline.

But as I read about disproportionate rates of suspension based on race, I also think about the increased use of School Resource Officers, which place a beat cop in many schools. This results in school discipline issues becoming criminal justice issues.

This all takes place against the backdrop of the growing pushback against Common Core aligned assessments. Getting into the full scope of these arguments is not the focus of this piece, but there are a range of organizations and groups actively opposing the state-level implementations of both PARCC and SBAC tests.

Frequently, opponents of standardized testing decry how these tests narrow the curriculum - and while that has some merit, the rationale provided by Xian Barret has more substance.

Which brings me to our assessment problem: standardized tests produced by outside organizations will not reflect or represent community norms - this should surprise absolutely no one. Teacher developed assessments are frequently held up as the antidote. But, I think about the discipline referrals leading to disproportionate rates of suspension among students of color, and that these discipline referrals started with teachers. The disproportionate rates of suspension in our schools point to larger bias issues within schools. It's hard to see how this entrenched bias would not seep into locally designed assessments.

, ,

Privacy Policies: Risk Mitigation and Strategy

3 min read

Privacy Policies and Terms of Service specify the rules under which an application can be used. They cover a lot of ground, and in this post, we are going to leave a lot out and focus on one specific element of the policies: how data collected by an application gets handled if the structure of the company changes.

In earlier posts, we have described how data collection within education is different than in consumer technology. In general terms, educational applications will require data to support a learner, and collect data about that learner as they work. The context within which that data is collected is pretty focused: it supports a learning process as defined by the software. In theory, the learner gets a benefit, and the data trail documents the process by which the learner earned the benefit.

If a company is sold, merges with another company, or goes out of business, the context within which data is collected and used has the potential to shift. Currently, most privacy policies treat a sale, a merger, or a bankruptcy as identical events, and most companies claim that learner data is an asset that can be sold as part of any of these deals. However, it's useful to look at how the context within which data can be collected or used can shift.

In many sales or mergers, the function of the application will remain largely unchanged. If two companies that offer online quizzing software merge, the data trails collected and stored pre-merge will be used in comparable ways post-merge. The same can often be true for a sale - there are many scenarios where one company could be acquired by another and the learning supported by the app continues uninterrupted.

However, bankruptcies are a different situation. In bankruptcies, the company is going out of business, or the company is getting broken up into smaller pieces. The original context within which data was originally collected no longer exists. In these situations, if data is treated as an asset to be sold, it can end up being used completely outside an educational context, in ways never intended by the learner. The main reason to reserve this right is as a hedge against failure: this allows investors to recoup losses on the backs of people who trusted the application.

Clear and direct Terms of Service and Privacy Policies make an application better. They eliminate ambiguity. However, there is no need for a strong service to claim rights over learner data if the original service is going away. If an edtech company is a service that supports learning, then the value is in the service, and the learning supported by that service - not in the ability to sell learner data. If, however, learner data is claimed as an asset, then the software is really a data collection tool with a specialized interface. While one purpose of privacy policies and terms of service is to define the rules that govern how an application is used, these policies also give us a glimpse into the larger strategies of the company.

,

A Question For All EdTech Companies

1 min read

I'm not one to bury the lede, so here's the question:

If you are a company that markets a product to learners, parents, teachers, schools, or districts, how does claiming the right to sell user data in case of bankruptcy improve your product?

Here's a related question for learners, parents, teachers, school decision makers, and district decision makers:

If you are using a service, how does the sale of your personal information improve the learning or experience you get from that service?

Many services and software applications marketed as education technology really function more as data collection tools with an interactive front end. One key distinguishing feature between a data collection tool and an actual edtech app is how the company pledges to treat your data. Privacy policies and terms of service provide the clearest view into the rights a company claims over your learning experience.

, ,

Two Steps That Would Improve EdTech Security Practice Immediately

2 min read

Natasha Singer has a couple of very strong articles highlighting weak or nonexistent security practices within many education technology (or EdTech) companies.

It's hard to say if this is just a poorly kept secret or the elephant in the room, but it's good to see the lax practices of EdTech companies getting more mainstream attention. It's needed - it's been needed for a while - and Singer's articles are part of the solution.

As I'm writing this, people are talking in Congress about the need for improved legislation to protect student privacy. And they are right, and I'm glad to see this being discussed in Congress - not because I think Congress will actually do anything, or that even if they did anything they would come close to getting it right - because it's good to see growing mainstream awareness that data privacy in EdTech needs to be fixed.

But there are two steps that could be taken almost immediately that would address a large number of sloppy security practices within EdTech.

  • Google and Apple both announce that they will no longer accept educational apps that don't encrypt data transfers. Existing apps will have a two month window to comply, or they will be removed from the store.
  • Venture Capital firms require security audits as part of their funding criteria. At a minimum, these audits would require encryption for both web and api access, encrypted data storage, and encrypted data transmission within their infrastructure. VC funding would be contingent upon meeting a base level of security.

It's worth highlighting here that some of the EdTech companies that have been rocked with security issues in the last year have received millions or tens of millions of dollars in VC funding. To put it another way: despite the fact that these companies were being handed large sums of money, no one anywhere in the vetting or due-diligence process saw fit to prioritize student privacy.

But with that said, taking either of the steps outlined above would cause an immediate improvement in security practices. If industry took the lead on both of these steps, the improvements would be immediate and industry-wide.

Then, we could focus on policies, pedagogical goals of the technology, and the need for learner control in shaping their learning. But getting the underlying security right is a necessary first step.

, , ,

What Opt Out Leaves Out - On Disaggregated Data

2 min read

The opt out movement is gaining steam. However, current opt out efforts often overlook or gloss over one essential element in the current educational landscape: disaggregated data at the school and district level.

The widespread use of disaggregated data was part of No Child Left Behind. The use of disaggregated data at the school level meant that schools now reported information on different populations within them. Rather than report just a schoolwide average, schools reported on specific populations, based on race and socioeconomic status. This brought increased focus - and in many cases, new focus - on how schools supported or didn't support different populations. The use of disaggregated data provided a starting point for discussing longstanding and long-ignored achievement gaps.

And, while you won't find many people arguing that NCLB was a stunning success, the emphasis on disaggregated data was a change that needed to happen, and that needs to be retained. Without the use of disaggregated data, we would not be having some of the critical conversations about race and socioeconomic status, and how the needs of populations have not been met in many of our schools.

When disaggregated data is left out of the conversations, that lands as, "we will go back to ignoring large numbers of school kids." When people within opt out talk about opting out of tests, though, they are often silent about disaggregated data. This omission isn't necessary, as there are many sources of data about schools, and school climate, and school performance, and alternative forms of assessment, that could be expressed as disaggregated data sets. Until opt out gets clear on the ongoing use of disaggregated data, the opt out message will alienate some potential allies.

, , ,

Scale for Engagement

2 min read

I've been thinking about the idea of scale, and how our current conceptions of scale are askew.

Within the Open Educational Resources space, there are a range of efforts that focus on broad scale adoption of OER. This vision includes hundreds or thousands of resources adopted hundreds of thousands to millions of times. In this vision, OERs replace textbooks, and save students hundreds to thousands of dollars on textbook costs. This is a very good thing. Arguably, this is a necessary step on the road to broader adoption of OER.

But, this vision of OER at scale is still predicated on scaling for delivery.

If we shift our vision of scale to tens of thousands of resources adopted tens of thousands of times, we begin to access some of the transformative power of OER. This vision of scale embeds and requires two elements: an informed and empowered teacher or learner, and high quality content available in a format that supports flexible remixing. These two elements - empowered learners and hassle-free remixing - begin to bring us to places you can't get to with traditional texts. Scaling in this way requires increased engagement, and it requires that we begin building system that remove barriers to engagement.

At best, current systems allow for flexibly copying content within a system - meaning that we have some flexibility within a circumscribed environment, but we lack the ability to work in our chosen space. While many platforms allow for export in a variety of open formats, every platform has context-specific display needs that are littered throughout the content. This cruft markup - injected everywhere, by every platform - is a huge barrier to both reuse and accessibility. If we are serious about scaling for engagement, we need to be serious about eliminating cruft content from canonical versions of openly licensed texts.

It's also worth highlighting that scaling for engagement would support scaling for delivery just as well. But we have a choice: do we want OER to just replace textbooks, or do we want open education to transform how we view teaching, learning, and authority? Scaling for engagement lets us do all of the above.

, ,

PPS School Climate Survey - Now With an Updated Privacy Policy

2 min read

On January 20th, 2015, Portland Public Schools launched a school climate survey as part of a first effort to get more feedback, more regularly from parents within the District.

This launch was a little rocky. We noted some problems with the privacy policy on the survey. The Oregonian announced the launch of the survey, but the way the Oregonian story described student participation would have had PPS running afoul of PPRA, and possibly COPPA (and to state the obvious, it would be good for the Oregonian to update their story to reflect reality).

Since the launch, however, there have been real and substantial improvements. PPS staff facilitated an introduction to the team at Oregon Kitchen's Table, the group helping PPS with administering the survey, and they were completely open to rewriting their privacy terms to make their intent clear, and to strengthen their privacy terms. Over the last few days, we all worked at editing the terms, and then met face to face to hash out the final version. Then, the terms were vetted by a lawyer who recommended some additional clarifications, and now, the updated terms are now live on their site.

Their new privacy policy is good - and that's not something I say often about privacy policies.

The folks at Oregon's Kitchen Table deserve a huge amount of credit for their responsiveness and work. They prioritized these improvements, and got them done incredibly quickly, without compromising anything in their mission, and without taking shortcuts. This is what success looks like.

, , ,

The Corinthian - ECMC Deal

1 min read

Many terms of service include provisions that allow for student data to be treated as an asset if or when an organization goes out of business or stops operations. The most visible example to date is arguably ConnectEDU. However, the sale of two campuses of the for-profit Corinthian College chain to ECMC Group, a student debt collection agency that operates as a non-profit, provides us another example.

Corinthian Colleges has been under investigation for predatory lending practices, and making false promises about job placements.

ECMC has also been accused of predatory practices when attempting to recover loans, and of raking in excessive fees off the backs of taxpayers and people who have defaulted on their loans.

In the Corninthian-ECMC deal, student data is getting sold from a predatory for-profit university into the hands of a predatory student loan collection agency. Companies win, students lose: the business of education, as usual.

,