New chapter on 'Small Data'

[Don Blair]

A few months back, Diego Rafael Canabarro had graciously invited Catherine D'Ignazio, Jeff Warren and me to contribute a chapter to 'GOVERNANÇA

DIGITAL' -- a new text out of UFGRS edited by Diego and by Marcelo Soares Pimenta. 

The result (thanks, Diego!) is here:

Chapter 6: "Less is More: The Role of Small Data in 21st Century Governance"

In this chapter -- inspired by Jeff's article from last year on 'Small Data', and incorporating insights from Ethan's blog post on recent citizen water monitoring efforts in China -- we pose a series of questions about 'Small Data': 

- What is 'Small Data'?

- Isn't it enough that my organization already makes data publicly available online?

- Can the public comprehend scientific results without special training?

- Is a bottom-up, grassroots, Small Data approach compatible with 'real science'?

- What about data validity?

- What does Small Data look like in practice?

- What about privacy and data ownership?

- Can my organization invest in both Big Data and Small Data?

- What problems might we usefully address with a Small Data approach?

- What are the challenges for the Small Data approach, going forward?

We then attempt to sketch out some answers, drawing upon our recent experiences working on the Open Water Project and citing examples from Dorn Cox's 'open source agriculture' projects in the FarmHack community.

Any feedback on the chapter -- or thoughts on any of the issues highlighted above -- would be very welcome!

Cheers,

dwb

[Don Blair]

@Mike -- your comments cut right to the heart of the matter. For all of our aspirations around facilitating a more 'grassroots, community-based science' approach, in which -- the fantasy goes -- the people who are most affected by environmental concerns -- who are, typically, located in communities with the least access to resources -- go on to employ tools (that they help to develop themselves) in order to collect data that they then interpret within their community ...  is this *actually happening*? Have we shown that it can work?  

My short answer I'd give to your question is: I think you're right; I'm not aware of very many examples of these sorts of 'community-led' monitoring projects being implemented and then assessed as 'successful' (or, even assessed at all) in terms of their impact -- especially for more complex sorts of environmental investigations.

The longer answer:  In the time since we'd written our chapter, those of us working on the Open Water Project have become increasingly concerned about this issue: the more we've learned about water quality monitoring, in particular, the more we've realized that it is a very complex subject, requiring a lot of auxiliary information and experience to get right.  We've begun wondering what sorts of social structures might best support useful flows of water quality information -- flows of data and interpretation within a community; flows that broadcast insights from the community to those outside; and flows of additional, alternative interpretations and analysis back into the community.  I'm very keen to look into the conference program you link to, and I'm excited to learn about Butler Universities' project around art and water (fantastic!);  if you learn of other such projects, please do share, down the line ..

- Recently, we'd been pointed to the work of researchers at Cornell that aims to address some of these issues; we found the following paper, in which Shirk et. al propose an overarching framework for constructing and assessing 'public participation in science' projects, extremely insightful:

Shirk, J. L., H. L. Ballard, C. C. Wilderman, T. Phillips, A. Wiggins, R. Jordan, E. McCallie, M. Minarchek, B. V. Lewenstein, M. E. Krasny, and R. Bonney. 2012. Public participation in scientific research: a framework for deliberate design. Ecology and Society 17(2): 29. [Link]

- Phil Silva (on this thread!) has been working hard to extend this approach and implement it 'on the ground' -- evaluating the impact and internal dynamics of community-based environmental monitoring projects.  He provides a great summary of this work in a blog post here (which also then links to the associated research article); I'd say that of all the projects I'm aware of, Phil's work is the most directly relevant to the questions you raise, Mike.

- I'm happy that Dorn Cox has already chimed in -- his work in setting up the infrastructure necessary to bring about this sort of truly grassroots set of information flows / loops, by focusing in on the context of small- to medium-scale farming, is truly visionary. (More details at http://farmhack.net);

- Catherine, Phil, Ethan, and other folks in the Public Lab community (and I) have recently put together an NSF proposal for an 'Environmental Storytelling Institute' that would focus on bringing more professional journalists into the mix; the idea being that journalists could play a vital role in facilitating the sorts of information flows among diverse groups that you've pointed to. 

- iseechange is a wonderful project, headed up by Julia Kumari Drapkin, and Lily Bui (on this thread!) that aims to facilitate dialogue among citizens and scientists around observed, on-the-ground weather extremes, historical climate data, and the content of a 'socially networked weather almanac';  it's already very exciting to see the ways in which farmers have already begun to blog observations about local conditions on their (beautiful) site; 

(@Lily, you can speak more to iseechange here?)

... and that's all I can write before getting out the door this morning.  Ah! I already see an email from Glorianna -- and gosh, yes, Living Observatory is an amazing example of another project that aims to address the issues you raised, Mike, by rendering the real-time dynamics of an an ecosystem accessible, virtually, to anyone with a browser  ... 

... please do chime in with any others I've missed, y'all?  I'm realizing that there are more folks on this list with their own, highly relevant wonderful, relevant projects than I can possibly account for -- and there has been an amazing curation of related projects on the sensor-journalism list in the last months -- perhaps we should gather these onto a wiki, as we shout them out?  We can easily create such a page on publiclab.org -- what might we best call it?  

Cheers!

Don

p.s. Oh, one more thing -- Ethan had pointed out that it'd be easier to share a non-PDF version of our chapter above ... not sure this is much better, but I've embedded the chapter in a research note, here

On Wed, Dec 3, 2014 at 8:26 AM, dorn cox <dorn...@gmail.com> wrote:

Hi Mike,

Great to e-meet you.  Your comments are right in line with the motivation behind the farm hack project, and an open source tool called FarmOS that we are building.  FarmOS is first and foremost a record keeping tool, but when linked with low cost sensors and used as a collaborative platform it will also provide decision support and adaptive management feedback that can ALSO feed environmental models.  We are in the process now of building an interactive soil health module for the Cornell soil health lab , and a grazing module.  Real world value generated by those who collect the data.  Would love to continue the dialog!

Dorn Cox

On Dec 3, 2014 6:14 AM, "Mike Barnett" <george....@bc.edu> wrote:

This is great.  However, the one thing that I would love to see is research and reports that address how and if the data is actually reaching the people who are affected by the data being collected (particularly those who are affected the most by what the data menas).  Everything I have seen is that the data collected by these open source projects is typically isolated to the groups that collect the data and only those who know about the project work, people with resources (usually more educated, higher income, etc..)… I’ve seen few projects where the data is shared and presented in any systematic way to lower income, communities of color, etc… 

 

There was a great conference last year http://www.nasonline.org/programs/sackler-colloquia/completed_colloquia/science-communication.html that explored how to communicate understandings of the data… and there isn’t much that targets the majority of the population (at least in the US).  there are some that are going… Butler Universities’ massive art/water project is one… but know of very few… would love to learn more… NSF is looking to fund some through the sciencelearning+ initiative but that is all that I know…  If there are projects out that doing this in a systematic way I would love to know them.

 

-Mike

 

 

Michael Barnett

2012 CASE/Carnegie Foundation for the Advancement of Teaching, Professor of the Year, Massachusetts
Professor of Science Education and Technology
Lynch School of Education, Room 123
Boston College

Like the Know Your Air project:  https://www.facebook.com/knowyourair

Like the Urban HydroFarmers project:  https://www.facebook.com/urbanhydrofarmers
http://urbaneco.bc.edu/mikeb/
barn...@bc.edu

781-367-2337

--
You received this message because you are subscribed to the Google Groups "Sensor Journalism" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sensor-journal...@googlegroups.com.
To post to this group, send email to sensor-j...@googlegroups.com.
Visit this group at http://groups.google.com/group/sensor-journalism.
To view this discussion on the web visit https://groups.google.com/d/msgid/sensor-journalism/CAA5k05TWGPeKC%2BP-jmdh3aFWw1Ar4EOV_8jygopvC0Y7UiSePg%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

[Gerald McCollam]

Don, Thanks for inviting feedback on these important themes. It’s unfortunate that a term like ‘big data’ should have stuck as resiliently as it has in the press and collective consciousness. In typical binary fashion we’re now compelled to counter ‘big’ with ‘small’ and argue for one over the other. Being ‘small’ naturally carries a connotation of being somehow less than enough, which I suggest is false in the context of science. On the one hand ‘bigger is better’ is not the case with scientific results or data in general. On the other our focus on scale obscures the more relevant issue which centers around how we determine relevance or salience in the data collected. How do we know when results are right-sized enough to bear weight? This is sometimes obvious (a picture of the crashing Hindenburg makes an effective argument for helium over hydrogen in balloons) but most of the time it’s not. Data very rarely comes bearing an interpretation on its sleeve. Still, some will argue that enough data alone can lead to discovery and that we no longer needs theories just lots more data. I'm not sure I agree with that. 

 

So, taking a first stab at ‘what is small data’ I propose first dropping the qualifier and focusing instead on how to determine when ‘enough data’ is appropriate to the task at hand. Sometimes small data – an old photograph found in an archive – is enough; other times many thousands of terabytes is insufficient.

 

Thanks for starting the conversation!

 

Gerald 

--
Post to this group at publicla...@googlegroups.com
 
Public Lab mailing lists (http://publiclab.org/lists) are great for discussion, but to get attribution, open source your work, and make it easy for others to find and cite your contributions, please publish your work at http://publiclab.org
---
You received this message because you are subscribed to the Google Groups "The Public Laboratory for Open Technology and Science" group.
To unsubscribe from this group and stop receiving emails from it, send an email to publiclaborato...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Gerald McCollam

cell: 347-739-0146

[Yagiz Sutcu]

That's a great discussion, thanks for starting it Don...

I personally do not like the term "big data" and its usage in everywhere because it's a hot buzzword right now...:) (https://twitter.com/schmittmaddy/status/453437767293624320)

What I wanted to say is, the data is important and essential but the information you extract from the data is much more important... So, if you can get the information you need (yes, what is information is a valid question too :( ) from the data you have, you are good to go...

Anyways, that is a great discussion... Looking forward to replies from everyone...

-yagiz

[Coby]

Also not sure I am a fan of the word small.  When I started the BeakGeek Project a few years back (since put on hold for various reasons) my vision was to start small to prove the concept, and then help build a network of bird feeders with sensors in order to increase the size of the dataset.  Idea being, if you could place enough sensors in schools and parks across the nation, you could not only monitor your local bird activity, but also start to see trends emerge in migratory patterns etc etc.  E.g. birds tagged with RFID chips and feeders equipped to read them.

Another crazy idea as example - Say you partnered with BASS tournament pros to place water quality sensors on their fishing boats.  They cover a lot of different waterways in a year, and collect a lot of sonar and mapping data already.  So even though it may be a relatively small group of people doing the collection the dataset could become rather "Big".  Especially if you engage the larger fishing community.

Your examples seem to focus on data collection in response to incidents that have already occurred rather than ongoing citizen science.  So perhaps Small Data works in this context and I am missing the point entirely.  It would make sense that the dataset be relatively small and focused since the incident in question impacts a local area.  If your applying citizen science to a local problem I could see the Small Data label making sense, but citizen science in general need not restrict itself to producing a small dataset.

Regards,

Coby

[wig zamore]

Yes, no reason small data cannot become big through aggregation and networking  For example, Weather Underground presented the possibility of integrating small air monitors at last year's EPA Sensor Conference in RTP NC.

Cheers, Wig

[Bryan Bonvallet]

It's interesting how people react to the phrase "Big Data." Since there have been a few reactions in this thread, I'd like to hop in.

Some Public Lab folks had written against Big Data awhile ago, but they were really addressing what I call Big Brother Data: the concept of government overseers tightly controlling large amounts of personal data. Thankfully I don't see the conflation occurring in this thread.

Big Data is an industry term that has leaked outside of the places where it a bit better defined. To address data versus information, Big Data was never about how to store a lot of data; in the industry, it is taken as a given that data storage is nearly infinite without any real challenge. Rather, how to process such large volumes of data into intelligible information was the challenge deserving of its own name. Big Data implies big information processing in the professional context.

The "big" in Big Data is a contraction of "so large our current technologies do not know how to process the volume in a timely manner." This is a moving goalpost. One of the original big data projects was the Human Genome Project. Here we had all these ACGT values ... but what do we do with them? How do we do it? Most of the analysis techniques of the time dealt with single files or what could be stored only in RAM, so say, up to 8 GB. New techniques were needed to process more information at a time and synthesize more processed information together in a mathematically coherent manner. That is what Big Data is about: pushing that boundary of "too large" into "easy enough", and then moving the goal post to the next "too large" increment.

So I hope that helps the understanding a little bit.

I'd say nothing we're doing right now is actually Big Data, because it falls within the boundaries of technological capability. 5 years ago, much of this citizen science data would absolutely be Big Data. Moving goal posts are hard to keep up for the media ;)

[Peter Lenz]

Working in Ad-tech we break things down this way:
Small Data - fits in a spreadsheet - excel
Medium Data fits in a normal RDBMS or NoSQL system (MYSQL/Mongo - includes sharding)
Big Data - requires Hadoop


Two points:
1) The vast majority of the projects will actually fall into Medium Data. Big Data is really, really, really, stonkingly hugely big, hundreds of thousand of data points per second minimum. Generally speaking if you think your problem is Big Data it's really just Medium Data being inefficiently collected. Consider if you really need all this data - usually you don't. Bayes theorem and downsamples are your friend.


2) Small Data is the ideal. It's human-scale data, something that can be fit into our heads. Anything larger then it is just about collecting more data and then reducing it using metrics and algorithms back into Small Data. It's complexity for complexities' sake and usually not needed.

[Jeffrey Warren]

Just a few thoughts:

> If your applying citizen science to a local problem I could see the Small Data label making sense, but citizen science in general need not restrict itself to producing a small dataset.

Actually i think that's what the intentionally provocative moniker "Small Data" means -- that a truly community-based science would be in the service of local issues -- and that "citizen science" efforts tend to miss that mark. The concept of "Small Data" as we're presenting it calls into question the goals of "general citizen science". However of course we acknowledge that it's not an either-or -- and Bryan's argued persuasively in the past that we're mainly talking about the purpose of the methods, not the methods themselves. I like the "Big Brother Data" idea too -- but highlighting the local scope of data by embracing "Small" celebrates the positive in our advocated approach, rather than sticking people in the eye; mainly a tactical/rhetorical decision. 

But yes, just to clarify, the idea is not that we're creating a taxonomy of the actual size of the dataset, but rather the purpose and premise of the work as situated in -- and run by -- a local community. 

Thanks for the feedback, folks! 

[Bryan Bonvallet]

Just going to +1 the breakdown of techniques.

If you have Big Data, then you are using something like Hadoop and you are very likely compromising on how you process the data even using the highest of processing technology like Hadoop. I think that's a pretty good up-to-date threshold.

I also agree very much that much of the analysis is about cleaning/filtering the data down to reasonable levels that can be better visualized and analyzed. Those problems also exist in big data, but are not exclusive problems to big data.

The real challenge posed by citizen science is not the same challenge of Big Data, but the rather the typical questions of experimental design and analysis: what data from this heap is important, what is the best analysis to extract the information I want, how can I be sure I'm not confounding variables, etc.
-Bryan

[wig zamore]

A few more small points in the QA QC realm.  As many of you know there is a big difference between citizen science that can be used for education and for relative, but not absolute analyses, and fully reliable data.  When it comes to using data as hard evidence for policy change, especially in the midst of adversity, there is a need to have very well calibrated instruments, or an ability to co-locate with same, or an ability for sheer numbers to make up for lower quality outliers, or some combination of the above.  And of course, to state the obvious, the ability to communicate findings clearly with a broader array of citizens drives the breadth of engagement possible.

Cheers, Wig

[Don Blair]

Hi All!

Fantastic discussion -- thanks so much to everyone for digging right in!  And now Jeff W. has nicely embedded the PDF in the associated research note, so it's an easy-peasy read: http://publiclab.org/notes/donblair/12-02-2014/less-is-more-the-role-of-small-data-for-governance-in-the-21st-century

To add to the discussion: the definitions we lay out in the chapter of 'Small Data' and 'Big Data' aren't really about the relative size of the datasets or particular algorithmic methodologies, per se; we're defining these catch-phrases in way that is intended to highlight the relationship that those who are conducting an investigation / developing a theory have to sources and production of the datasets.  These are, as Jeff says, an intentionally 'provocative' set of definitions; in case it's of interest, we sketch these ideas out in the first few pages of our chapter. All that said, I'm finding the discussions here (using a range of notions of 'Big' and 'Small' data) very useful & thought-provoking.

Cheers!

Don

On Wed, Dec 3, 2014 at 1:55 PM, Coby <cgl...@gmail.com> wrote:

[Coby]

"Generally speaking if you think your problem is Big Data it's really just Medium Data being inefficiently collected."

Ha, that's perfect.  Part of the reason I put BeakGeek on hold, besides my complete lack of funding and availability of dirt cheap sensors, was how to handle images of birds from feeder cameras.  Lot's of repeat visits by same birds among other issues.  But mostly, the tech needed to have the server(s) process 1000's of images every day to help ID  species etc.  Perhaps even track specific birds that had distinct markings.  Nothing on the Big Data scale, but a huge problem in scalability for me none the less.

@Jeffrey I get the provocative part of calling it "Small Data" - Still don't think I am sold.  I like Local Data or Localized Intelligence better, but it semantics.  Besides it all gets sucked up into the Big Data vacuum at some point.

R,

C

[geraldmc]

Ultimately I think we’re focused on what's meaningful and transferrable from person to person, whatever the volume or type of data it is. I agree that that meaning is best expressed or is most potent at the local level. I also agree with Wig that 'small' data can easily morph to big depending on the use case, so it remains to my mind not a useful distinction beyond whatever rhetorical value it may have. 

Whether it’s stinkingly large or pathetically small ultimately what’s of value is the meaning that’s revealed, why it’s meaningful, and for whom. In the end perhaps all meaningful data is small, in the same way that all politics is local. 

Great stuff Don, thanks again!

Gerald

[geraldmc]

Isn't it enough that my organization already makes data publicly available online?

Most open data is ‘accessible’ by downloading in some form (as csv or db file) or through an application programming interface. API access works through a client app that makes request and response handling more efficient. All in all I’d say it takes some degree of expertise and commitment to make actionable use of most open data.

Socrata makes an API for government data that works quite well. I’ve seen it in use in a number of municipalities, including New Orleans. I think it’s fair to say that most use of APIs like Socrata stays within the context of the site that hosts the service, i.e. in NOLA it remains buried within the local government's website and is dependent on their continued support and interest. There’s not much outside repurposing of data by community-based services here, which is kind of the whole point. How open data is to be ‘rendered legible, and meaningful, for the various public audiences’ is up for grabs. In practice it's left up to the community to figure it out. 

-g

p.s. Don, this time I read the whole article before commenting on it ;)

On Wednesday, December 3, 2014 4:44:15 AM UTC-6, Don Blair wrote:

[George Gallant]

I think the terms "Big Data" and "Little Data" are being taken
completely out of context. They really apply to an upcomming Star Trek
movie in which Commander Data finds true love and has an offspring. The
new "Data" is 1/8 the size of the size of the original and was going to
be named "Bit" to reflect his parents names of "Byte" and "Octet" but
the military frowns on using first name whereas action monikers are
acceptable.

[Coby]

I think this point works well with how the chapter concludes with "deeply optimistic".  Despite my issue with "Small", that's a tiny point to make in the Big Picture.  OK, enough about relative size.  I can remember ten years ago when just about all gov data was either not shared, shared but completely unstructured via 100mb ppt or pdf's, or structured using some proprietary software.  Thanks in no small part to folks like Public Labs, Crisis Mappers, and others in the open source and open data movement there may be light at the end of the tunnel.  So despite me being jaded on the issue (I gave up on open data fusion in the gov sector after a few years of explaining Twitter) I do think there is reason to be optimistic.  

To the point of the chapter.  It may be important to make a distinction between Citizen Sensors and Citizen Scientists.  I believe there are people who have interest in helping by providing sensing/collection and the results of said sensing/collection, but have little/no interest in the scientific method in between. E.g. Fishermen placing water quality sensors on their boats and taxi/FedEx/UPS/Uber drivers placing air quality sensors on their cars.  They provide the mobility, infrastructure and possibly even monetary support, but they may not want to provide much science.  So I think there are levels of cooperation and interest in the Small Data process/cycle/spectrum that could be leveraged without every person/organization on the team directly involved in the science/geekiness.  I think the later may be achievable as a long term goal when tied to the cool STEM education stuff that people/orgs like PLOTs are supporting.  But perhaps, in the short term, Small Data gets more traction by finding very low friction ways more people can become involved.

R,

C 

[wig zamore]

Coby,

That all sounds just about right to me.  

On the academic Big Data side, there are petabyte quasi-public data sets being shared in the genetics and epigenetics (more interesting) world.  Broad may do 50,000 or so whole genomes in the next year +/-.  The non-coding RNAs (micro and long) thought to be just junk a few years ago are exploding the complexities and also the insights possible into environment - lifestyle - biological evolution and response.  

On another path, Public Labs and similarly interesting grass roots entities, will likely continue to converge with the exploding ability for citizens to map noise and sound, other environmental factors, diet, exercise, ventilation and metabolic rates, etc.  The increasingy common personal electronics tools with space and time stamps are an incredibly powerful base technology for citizen science, engagement and response.

The citizen science and engagement is scalable.

Cheers, Wig