Structured Data for Humanitarian Technologies

Lunch break in the sun :-)

Lunch break in the sun 🙂

The Association for the Advancement of Artificial Intelligence (AAAI) organises every year the Spring Symposium Series (SSS) at Stanford. This is an opportunity to look into new topics for AI through a set of parallel workshops lasting 2 and a half day. After having attended this event for the first time in 2010, to speak about computational intelligence for the semantic web back then, I got the chance to attend the 2015 edition. This time I joined a workshop about structured data for humanitarian response to present our work on the Entity Registry System (ERS) and discuss its potential usage in the humanitarian context. It was a very lively and friendly workshop with a relatively small crowd of 12 persons exchanging ideas and experience in an informal setting. The organisers put together a nice program with 4 paper presentations, 3 (amazing !) keynotes and 2 panels. In what follows I will try to summarise the main points I took away from all that.

There are societal divides

The digital divide is something very much spoken of but there are also gaps between societal groups. Namely,

  • Between people affected and those here to help: as matter of fact, the very first respondents to a crisis are not the NGOs or other big official bodies. It is the neighbours, the friends, the family that are the first ones to respond. They are also the ones having all the local knowledge. Everyone willing to help should aim at keeping the local people “in the loop” and stay as close to them as possible. Focusing on questions such as “what data do they need ? what could make their day better ?” can help shaping priorities. Including them in the data collection process, and feeding them back with the results of data analysis can help building trust and increases the effectiveness on the intervention.
  • Between those on the field and those working remotely: those working on the field can found themselves spending time “feeding the beast”, sending reports and data to those working remotely in organisation head-quarters or elsewhere. This is a necessary activity in order to those working remotely to get a grasp of what is actually going on on the field but one has to keep in mind that time spent reporting is time not spent helping people in some other concrete ways. This gap between local and remote work can be tackled keeping in mind questions such as “Are the maps produced of any use locally ?” or “Will this new reporting tool slow down the activities of those having to use it ?”.

ICT is not the issue, probably nor the solution

Really. Think twice before getting on coding something. People doing intervention are typically not geeks. ICT is a great tool for improving the speed and scale of data/information/knowledge flows but in a context where resources are scarce the focus must be on simplicity and efficiency. The favourite data tool of crisis responders is a piece of paper and a pen. This is a proven reliable technology which is usable in direct sunlight, cope with dusty conditions and does not run out of batteries. If an information has to go from A to B someone will always figure out of having this happening, with or without any technology involved. Thus, when data does not fly from A to B the problem is most probably not the lack of ICT.t

If ICT is to be used this has to be done in an homoeopathic way by small influx in parts of the data-driven process rather than as a big wave of changes establishing brand new practices overnight. Every solution designed for the responders should actually be co-designed with them in order to increase the odds of seeing a good adoption rate. This is something we successfully applied in our work on voice-based services for African farmers. Besides, it is also important to acknowledge that machines and humans are both good at some things and less good at others. Finding the sweet spot for splitting the tasks among these two groups is key to an efficient and effective usage of ICTs. Besides the collaboration between these two groups (think graphical user interfaces, vocal interfaces, …) the intra-group collaboration among machines and among humans also needs some care. For the former it is the definition of common APIs and standardised vocabularies/thesaurus that will enable friction-less data sharing among different computing tools. Some example of this are MOAC, HXL and the facility registry. There is more work to be done there and standards such as Linked Data Platform (LDP) could most likely be put into good use. The later, communication among humans, brings its own set of challenges. This will have to be shaped around data re-use policies, agreements and trust. It becomes increasingly visible that Big Data as, e.g., the data telecom operators have can provide valuable insights onto population dynamics. But as the impact of these insights will be depending on the aims of people having access to them it is important to keep the data flow open but nonetheless in control. One way into this could be to establish some kind of cross-organisational agreements to share “key” datasets among each other when this becomes necessary. A similar idea was also proposed during the discussions at the event “Responsible data for humanitarian response” some weeks ago in Den Haag.

Get prepared and be flexible

In WWSW and W4RA we commonly deal with ICT4D use-cases for which we have time to experiment and can afford system failures from time to time. In sharp contrast to this speed and efficiency are key to interventions related to humanitarian crisis. There is little, if any, room for experimentation and no time to innovate. Quoting Eduardo, “Better fail in a traditional way rather than failing trying to innovate in the middle of a crisis”. This means that all innovation and experimentation work must happen before a crisis shows up. In many cases this will also implies doing the R&D work using budget allocated for other research activities or having no budget at all as this work will necessarily happen before massive resources get allocated for dealing with the crisis. Building resilience is thus an important task that has to find its driving need outside of a crisis context. Projects such as 100 resilient cities are highlighting this. As part of finding a driving force outside of a crisis, one also has to accept and anticipate that, e.g., a resilient SMS system for alerts will also be used to send seasonal greetings at the end of the year. This is not a bug. It’s a feature showing that people really embedded the tool in their daily life and will thus be more likely to make use of shall a crisis happen.

Some more random thoughts

  • It feels a bit weird returning to a place after 5 years and see that absolutely nothing has changed. SSS15 was exactly like SSS10 on all aspects except the content 😉
  • Coffee was good and was always served in big cups
  • Humanitarian response is a very large domain with a lot of professional organisations, software and people involved. As showed by Sara Terp, there is a large crowd willing to help
  • The suite of tools built by InSTEDD is truly impressive !
  • We should keep an eye on the ISCRAM conferences
Advertisements

I am a researcher mainly interested in : architectures for publishing, consuming and preserving Linked Open Data in low-resource contexts; complex systems; education; data visualisation; video games

Tagged with: , , ,
Posted in Events
%d bloggers like this: