Thursday, 27 September 2018

We are both underestimating and overestimating the dynamics of Artificial Intelligence for KM: My 2 cents featured in the updated Agenda Knowledge for Development (K4D)

Today the Knowledge for Development Partnership (K4D) initiative released an updated version of the 'Agenda Knowledge for Development', which formulates 14 "knowledge goals" meant to supplement the UN's Sustainable Development Goals and strengthen the Agenda 2030 from a Knowledge Management perspective.

The updated version includes one added Goal 14 ‘The arts and culture are central to knowledge societies', as well as 57 added statements (in addition to the 73 statements in the previous edition) from knowledge management practitioners in development across the world, including my own.

I've already shared on this blog my reflections on the role that I believe Artificial Intelligence will play in Knowledge Management in the coming years, and I've re-iterated my view on this in my statement for K4D which is now featured under Part II of this document (page 84), "Statements on Knowledge for Development" next to the statements from 130 of my fellow KM colleagues:

"It is easy to hail knowledge as ultimate driver for change and key resource to achieve the SDGs. The sobering reality, however, is that more knowledge doesn’t per se make the world a better place. In fact, one can argue that humankind has reached a point in history where there is more knowledge than it can productively handle. Despite the known benefits of democracy, support for democratic principles is shrinking worldwide. Despite the advances to human progress through science, increasing portions of populations wilfully choose ignorance and ideology over scientific evidence. And despite unprecedented access to news and information sources, consumers chose to rely on fake news instead of fact checking. These are symptoms of a world in which there is just too much information for the human brain to meaningfully process. And the instinctive response is to retreat to what we already know and are comfortable with, rather than expose ourselves continuously to a complex world in which discerning the best route of action among many truths is very hard work and just plain exhausting.

One way in which humans will try to resolve this in the next decade is that we will turn to Artificial
Intelligence (AI) to sift through the massive amounts of knowledge and information available, and make sense of it for us. As with past tech trends, we are currently both underestimating and overestimating the dynamics of this technology in the way we manage knowledge. We are underestimating the profound transformational impact AI will have on the way we learn about, curate and analyze examples and insights from worldwide activities in our everyday work. And we are at the same time overestimating the extent to which technology can solve our underlying problem of using knowledge to better the human condition. Programmed biases in AI systems, questions of legitimacy and over-reliance on ‘black box’ AIs, and issues around ethics and local context are just some of the problems that we will have to resolve as we will increasingly rely on machine learning. Knowledge for development needs to be mindful of the issues that knowledge complexity is triggering in societies, and brace itself for the full force of the AI revolution that will transform the way we manage this knowledge in the upcoming 10-15 years so that we, as development practitioners, are well positioned to both reap its benefits and mitigate its pitfalls as we work towards achieving the SDGs."

What do you think? Is this view too much preoccupied with the current global political context, which may be more of a momentary snapshot than a long trend, or do you agree that the dichonomy between information availability and capacity to dicern and process it will be the key KM challenge of the years to come?

Monday, 26 February 2018

How to program for uncertain results? The innovation journey of a 'slightly unsusual' programme in UNDP

Innovations are driven by risk-takers. Part of UNDP’s role in innovation is to provide the space for risk-takers to develop and test their ideas. And it turns out that sometimes these are not individuals, but entire programmes! The Pacific Risk Resilience Programme (PRRP) covers the Pacific countries of Fiji, Solomon Islands, Tonga and Vanuatu, and takes an unusual approach within UNDP’s programme portfolio. Now in its fourth year, it didn’t exactly follow the standard programming approach in which a challenge and a development model is identified in the beginning, and a set of interventions is designed that would then be rolled out over the course of the next years, with clear activities and results prescribed for each year. Instead, PRRP didn’t actually describe the model or the interventions themselves at all. Instead, they let the model emerge over time by running sprints of interventions and evaluating them frequently, something that is known in the information technology world as ‘agile development’. I've talked to the programme manager, Moortaza Jiwanji, about their approach, what they learned from doing things differently, and the implications of their experience for UNDP programming.

Q: Why did you feel the need to do things differently than in ‘traditional’ programming?

The main reason was that we that we had to develop something for which there was no precedent. We were venturing into unknown territory, and it made little sense for us to prescribe what results would look like 4-years in advance with a results framework that pretends to know exactly what activity would be best to deliver by year 4. We simply couldn’t see how that would work.

Climate change and disasters have a real impact on people in the Pacific. Despite the unprecedented levels of funding and programming in the region, it is disheartening to see how communities are still experiencing the same types of impacts of climate change and disasters and in some cases, these are becoming worse! This is particularly concerning given that the symptoms of climate change such as cyclones, flooding and droughts are likely to increase in intensity and frequency in the future.

Much of this programming in the Pacific has led to some concrete results on the ground, but we felt that there was not enough thought being put into addressing the root causes of these vulnerabilities. It is less obvious to see how development itself is being adjusted to address these risks. For instance, why is it that schools and houses are still built in flood prone areas without the appropriate materials and design codes? It is also becoming increasingly clear that development itself is a primary cause of this vulnerability to climate change. This could perhaps explain the cyclical nature of these impacts.

We realized that something needed to change within development itself and not just in climate change programming. Most programming is focused on technical solutions such as building sea-walls rather than dealing with the root causes. It seemed at the time that there was not much programming experience in dealing with climate change from a ‘development’ perspective. We knew what needed to change but there was limited experience in the region and globally to show us how.

That’s why we decided to develop a model for risk-informed development without any preconceived ideas about what it would look like and how it would work. Our starting point was to address deep-seated governance issues, not for climate change, but for development. We also felt that this would be an opportunity for UNDP to build a niche for itself, particularly given that we are a ‘development’ agency and also deal with governance reform. We were able to do this through the Pacific Risk Resilience Programme (PRRP) which started in 2013. PRRP was funded by the Australian government as they also felt willing to try something different given that they were also not seeing the aggregate results in the region.

Q: What exactly was ‘different’ about your approach?

We knew we had to do two things differently: First, in order to tackle the root causes of climate change and disaster risk, we had to work deep within development itself. And second, because at the time back in 2013 there was not much experience of dealing with climate and disaster risk from a truly ‘development’ perspective, we had to follow an approach that was largely experimental at the time and depart from more traditional approaches to programme design and implementation.

So what was different about our approach? First, unlike most development partners in this area we did not work as an outside partner with climate change and disaster management functions in government. Instead we programmed ‘from within’ governance systems where our government partners owned the development interventions from community to national level. We also used a human-centered design approach that really focused on developing individual mechanisms with the same people that were going to apply them in their government ministries and agencies. Both these aspects allowed our country partners to help design and fully lead the initiatives themselves rather than UNDP leading the way. This admittedly raised some eyebrows at the time as there was an expectation for climate change to work with (and provide funding for) the ‘usual suspects’.
Figure 1: The Innovation Feedback Loop

Second, unlike standard programme design approaches we did not predetermine our activities and outputs well in advance for the next four years. Instead we built smaller and targeted experimental interventions where we saw some prospects of success (see adjacent diagram), e.g. where we found receptive partners and conducive political environments. This allowed us to understand which activities did yield the best results as we implemented them. We then spent a great deal of time and energy in measuring any apparent successes and even more importantly failures. Learning from these experiences was the most important ingredient and we spent a great deal of time and energy on doing this collectively with our partners. So whether a pilot leads to measurable successes or failures, the real success of this approach comes in how well you learn and subsequently redesign and modify approaches based on these learnings. Based on this iterative process we would then develop the overarching model as it emergedfrom those experiences. The interesting thing about this experience is that we developed this modus operandi ourselves, completely unaware that UNDP was promoting through its global Innovation Facility exactly such innovation and design approaches that encourage this type of experimentation. At the time of designing this programme back in 2013 we decided to call this approach ‘emergent design’, but it aligns very much with the innovation principles of agile development and problem-driven adaptive iteration.

Based on the learning from these experiments, we have now developed an overarching model around the concept of ‘risk governance’, designed and tested to risk-inform development ‘from within’ and at all levels of governance. For more information you can read our recently launched policy brief and you can also see practical examples of how this is benefiting countries in the Pacific on our website

Q: What were the challenges you encountered?

Risky business. Developing a programme based on emergent design or agile development principles is extremely exciting. However, it can also be quite stressful because in essence you are taking a significant risk in programming something that has not been tested successfully yet. This is particularly challenging when it comes to convincing your programme stakeholders, such as your donor, country partners and even internal management.

Raising eyebrows. In the early days, we seemed to develop somewhat of a reputation as being the slightly unusual programme within UNDP. This was not always cast in a positive light and this is partly because we did not have a fixed and clearly defined results and resources framework over a four-year period.

Buy-in from stakeholders. There are three types of stakeholders that we dealt with through this experience: the country partners (or beneficiaries); our donor partner; and UNDP itself. The approach of working from within and building governance systems to risk-inform development was most positively received by our government and donor partners and then eventually with our managers within UNDP. This took a little while perhaps largely because we were venturing into the unknown and did not have a clear narrative to describe and justify our approach, particularly in the early days.

It can take some time to show predictable and regular results. Agile development or emergent design approaches can take some time to achieve tangible results. It is almost by definition impossible to predict when and how results are going to be achieved. This was particularly challenging when working in an environment where programmes are expected to report on results against clearly defined outputs and targets at least every quarter.

Q: What were the benefits of taking this approach compared to more traditional approaches?

In essence, agile development allowed us to get results that otherwise would have never emerged had we prescribed our specific outputs the traditional way several years in advance. And on top of it, the solutions that we did get through the agile development approach now address the actual problem we’re trying to tackle much better.

Over time we saw that taking this approach was extremely beneficial, particularly to our government partners, in offering more sustainable and realistic solutions to the complexities of climate change and disasters in the Pacific. You can see this by the fact that our country partners are now advocating for this approach within their own countries.

Unexpected solutions. What’s really interesting is that taking this approach has led to solutions that we would have never designed up front. For instance, we now have Ministries of Women leading on climate-informing community development initiatives. Private sector networks are now being formed to not only work better together in times of disasters but also to provide a more effective link with government and partners. Local governments are leading the way in risk-informing infrastructure projects. You can see these examples and others on our website under ‘Results’ on

Ability to adapt. Secondly, most of our country partners have really appreciated the ability of UNDP to adapt to a constantly changing environment. They often feel that projects that have fixed activities and outputs for a four-year time horizon are unrealistic and can compromise their own ability to initiate real change on the ground. What we see now is that our partners are leading the way and collectively we continue to discover new innovations.

Finally, taking this approach to development programming is immensely rewarding both on a professional and personal level. It almost feels as if there is no other way to deal with the complexities of development in the Pacific and even beyond.

Q: What would you recommend to others who want to take this approach?

I would recommend four key things. First, don’t be afraid to fail and be completely open about this to your partners. This is critical in finding innovative solutions to complex development challenges. Secondly, invest in smaller and manageable initiatives through prototypes. This will help minimize your risks and allow for real creativity. Third, you will have to tailor your results framework in a way that you frame your described activities and outputs as e.g. number of experiments run and evaluated, number of experiments identified for scaling up, etc., rather that describing up front what these experimental interventions will specifically look like. This will give you the leeway to explore uncommon and innovative solutions, while at the same time hold yourself accountable to measurable milestones within this agile development journey. Finally, taking a leap into the unknown can be risky and cause negative perceptions of your work around you. Develop a small group of like-minded colleagues from within and outside the organization who are genuinely willing to try this out and support you. At the same time, it is imperative to engage management early on in an open but confident way about what you are doing and why.

Q: What could all this mean for the future of UNDP’s programming?

We had very interesting conversations with counterparts within institutional donor organizations who frankly told us that refining this agile development approach further could be very rewarding for UNDP. It would allow the organization to position itself as a unique implementing partner that can offer a different way of programming than most other implementation contractors, especially in programmes that try to tackle government reform issues. I feel that the future for UNDP and similar organisations working in this space lies in innovating its programming itself through such agile development, or ‘emergent design’ principles. Not exclusively, but at least as part of its portfolio. Not only is there a lack of this approach in the development space, but more country partners will want this because it is particularly suited to addresses complex development challenges for which no clear solutions exist yet. This needs to go beyond mimicry though and requires fundamental behavioral shifts in terms of how we design, execute and evaluate our work. But the outcomes are worth it. As I said, this has been the most rewarding professional and personal experience for me so far.

Tuesday, 16 January 2018

Artificial Intelligence will change Knowledge Management as we know it

I recently came across this blog post by a start-up that is developing an Artificial Intelligence that is being trained to read and write at the level of a specialized human analyst and produce briefings in human language based on a set of different information resources. It’s just one example of many different companies that are currently working on this challenge. The obvious clients are intelligence agencies, governments, or news agencies, but eventually this will enter all of our everyday work very soon.

I thoroughly believe that this is what knowledge management in large organizations will look like in 10-15 years from now. In my organization, we’re challenged daily to consolidate the key lessons and insights from all our country-level programmes and experiences, lest meaningfully combine them with information, trends and insights from the larger development sector. We complain that we’re overwhelmed by the information overload that social media, Yammer and knowledge networks impose on us, and retreat to focusing on a narrow set of information that confirms our biases, pretending we know what we need to know, when in fact we always only have a small piece of the puzzle. Artificial Intelligence promises to overcome this dilemma, as it will have immediate access to all information available, and can do the necessary analysis for us.

We might not be quite there yet to make this practical for organizations like UNDP, but we’re getting closer and closer. Last year we as UNDP KM team at HQ engaged with a well-known AI systems provider, and while both organizations were not quite ready yet to commit partnering on an AI system that can make sense of unstructured texts, trends, insights and lessons in the development sector, will have to get real about this soon if we as an organization want to be ready for what is to come. To quote the same article above: “With technology that can read and write, you have the flexibility to generate custom insights in any format or level of detail. If you’re a subject matter expert, Primer can tell you a detailed story that takes your knowledge into account. If you’re new to a subject, it can generate an introduction to get you up to speed quickly. If you have an interest in a particular angle on the story, or a geographic lens that you want to zoom in on, the insight can be customized for you. Imagine the possibilities if you had one thousand analysts working for you, all day, every single day. What questions would you ask, what kinds of briefings would you have them prepare?”

Now, technology can always only be one part of the solution. It is important to keep in mind Dave Snowden’s adage that if you have $1 to invest in KM, invest 99 cents in connecting your employees over shared opportunities and 1 cent on content. Connecting people has always been (and will always be) at the center of knowledge management, where we try to connect staff to those who have the skills, capacitating them to identify the right people, enable them to collaborate and research in real time and turn the result of that into actionable insights. It’s why other UN organizations have often looked to UNDP for KM advice, because it regularly chose to make strategic investments in connecting and fostering networking among its staff, being the first UN agency to pioneer email-based knowledge networks in 1999, the first to introduce organization-wide corporate social networking with its award-winning platform Teamworks in 2009, and continuing that trajectory with Yammer (among other things) today. Connected people are the ‘operating system’ of any meaningful KM effort that allows real-time collaboration within a human context that can be turned into actionable insights.

But what forefront thinkers in the AI space tell us about AI’s implication for governments is true for international organizations like UNDP as well: AI will relieve knowledge workers form drudgery tasks, split up our work into automated tasks (e.g. research, collation) and human tasks (value-based decision making, social interactions), and augment the capacities of knowledge workers by adding layers of real-time and predictive analysis that humans couldn’t do by themselves. Together with many of my KM colleagues who are much more skeptical about AI than I am, I also believe the focus will be on augmentation, not replacement. Nonetheless, all indications suggest that we are at the beginning of a revolution of how knowledge work looks like, and organizations like UNDP will be affected internally by both the benefits and risks. The only way to get ourselves ready for it, is by doing what the innovation community always does: Striving to get our feet wet early, and learn, learn, learn.