Data and Development in 2030: Six predictions for the upcoming decade

It is 2020, the beginning of a new decade and an appropriate time to make predictions about what the next 10 years will bring. We recently posted an article about the “all data revolution” and how the way we work with data is undergoing fundamental changes across the data and policy cycle. But, what does that mean concretely? How will our work look different 10 years from now, given this revolution?

Any predictions about the next 10 years must take fundamental trends into account. These trends provide the backdrop against which the next decade will unfold. Helpfully, our colleagues at Oxfam have recently summarised these ‘megatrends’ in a post here. In essence, there are four key trends: First, the climate emergency will increase in severity. This will have knock-on effects on all kinds of other issues such as migration, biodiversity, food security, conflict, economic growth, economic policies, and others. Second, the demographic composition of the world’s population will change with an ever increasing proportion of youth living in Africa – in particular in urban areas. Third, modern digital technology will develop further in terms of reach, i.e. more people will gain access to the internet, in terms of usefulness, e.g. with improved AI, and in terms of ease of use, e.g. with algorithms deployed seamlessly to our smartphones. Fourth, geopolitics will see a struggle for shifting power centres, mainly between the US and China.

These developments will shape policy-making across the world in the next decade, including areas specifically within the development sector and amongst stakeholders. What will these advances mean for those of us who work with data in development? Here are OPM’s six predictions on what will look different in 2030:  

1. Deep Learning and AI will feature in our day-to-day work. Richer personal data will be available across the world, including in developing countries (remember – Africa’s large young population will have increased access to ever cheaper smartphones) and therefore  sophisticated AI will be much easier and cheaper to deploy. This means that everyone, from government officials to NGO front-line workers, will have some sort of algorithm in their pockets, or computers in the office, that will make their lives easier. Existing examples include computer vision algorithms to identify child malnutrition, machine learning methods to automate scanning and processing of paper-based CVs, or AI that can analyse photographs to identify crop diseases. Over the next 10 years, these approaches are going to be trialled, tested, and deployed with increasing frequency and usefulness.
2. Surveys won’t go away, but they will regularly be combined with ‘big’ data. As I have previously argued, purposefully collected and sampling-based survey data will continue to be important as they provide information on populations with a degree of reliability and known uncertainty that other data sources do not. However, this data will increasingly be combined in smart ways with data from other more unconventional sources – e.g. satellite imagery, crowd-sourced data, mobile phone records, remote sensing – to create ‘data sandwiches'* that allow to produce estimates of key indicators at higher resolution and higher frequency than survey data alone. Studies have shown how this can be done e.g. for poverty, population numbers, or other development indicators. These approaches will become easier to apply and access, and resulting estimates become more commonly used to inform policy planning and implementation. In 2030, we won’t use outdated data or wait for the next Demographic and Health Survey (DHS) or census to come along in five years to inform activities or measure SDG progress – we will use estimates derived from data sandwiches instead.
3. Text analytics will become key to monitor reactions to policies and interventions. An increase in access and use of social media will mean that populations will express their opinions online more readily. Better off-the-shelf algorithms to implement text-as-data analyses will imply that this increased chatter can be readily analysed to inform policy-makers and practitioners about sentiments of individuals or populations towards their activities. In rich economies and universities, this has already been happening for a while, but the next 10 years we’ll see these approaches spread beyond academia and into the non-OECD world as well.
4. The use of mobile phone and satellite data to monitor climate change, disaster risk, and related effects on populations in developing countries will significantly increase. This is the result of a confluence of different factors: first, as in other areas, there is now extensive academic literature on how to do this (e.g. see here, here, and here). Second, thanks to the work of organisations like flowminder.org and UN Global Pulse, the data and tools needed to do this are becoming easier to access and use. Third, as mentioned above, the demand for these analyses will increase given the severity of the climate emergency.
5. There will be a worldwide scramble for data and AI hegemony, including in Africa. Much has been said about how the power struggle between China and the US is playing out in the sphere of digital technology (see here, here, and here). Increased Internet usage and availability of data in Sub-Saharan Africa will mean that there will also be increased demand for tools to derive value from this. The two main geopolitical players in this field will try to offer and sell their algorithms, approaches, and analyses into governments and the private sector on the continent.
6. Data protection, privacy, and the ethics of AI will become of central importance to our work. Even though much has been written about data protection and the ethics of experimenting with big data, AI, and machine learning in development, the sector has yet to agree on a widely used or recognised governance framework and remains extractive in nature. While much of the data for development work so far has been implemented under the ‘do good’ or ‘data philanthropy’ label, for some this amounts to a new ‘algorithmic colonization’ and exposes the sector to a serious risk of misuse of data. Hence, there is a high probability of a ‘Cambridge Analytica’ type scandal happening in international development. The hope is that this will push stakeholders to more responsible behaviour and to an acceptance that issues around data protection and ethics of AI are crucial to actually ‘do good’ with data in development.  

Together, these predictions point to an exciting decade ahead of us. On the one hand, many of the predictions above build on the idea that experimentation and research implemented in the 2000’s or 2010’s will be mainstreamed, standardised, and scaled up across the world. This bodes well for a more effective and efficient monitoring of the SDGs. It might also make achieving these goals more likely. On the other hand, the two last predictions point to the quite significant risks for conflict and ethical issues derailing the efforts to use digital technology and data for good. It will be the job of researchers, government officials, and practitioners in the sector to ensure that by 2030 the benefits will prevail.

*Full disclosure: credits for this term to Prof. Andrew Tatem at Southampton University. He first mentioned this at a seminar at OPM.