Algorithms are becoming an increasingly pervasive aspect of everyday life – so much so that a recent UK Council for Science and Technology report described our epoch as the ‘age of algorithms’. Yet despite their prevalence, often providing the driving force behind our usage of a range of products and services from mobile phones to insurance policies, they remain relatively obscure to the general public. This is primarily down to their auxiliary role, unlocking meaning from data and underpinning the technology that drives much of everyday existence today.
If we are still in the throes of what has been called a ‘data revolution’ then it logically follows that algorithms will be of increasing importance. How though, do they relate to development?
By George Bodie
Perhaps the most obvious example here is in medicine. Often used by clinicians for outcome predictions or for the choice of treatments, algorithms are becoming increasingly important regarding standardised care procedures – in the UK almost all primary care trusts now have standardised algorithms for the choice of medications for patients with psychiatric conditions, for example. In Africa, algorithms are being developed for malaria-affected regions which aim at increasing the specificity of malarial diagnoses in order to better distribute anti-malarial drugs.
In response to the recent outbreak of Ebola, the Canadian company Chematria programmed supercomputers with an algorithm which simulates and analyses millions of potential medicines to predict their effectiveness against the disease – a process they claim can replace physical lab testing, giving huge returns in speed and cost.
Algorithms have even been developed for atrocity prevention – with a recent competition run by the US Agency for International Development (USAID) and NGO Humanity International giving sums of up to US$12,000 to develop and pilot algorithms which use socio-political indicators and data on past atrocities to attempt to predict future ones.
Algorithms have an important role to play beyond life-saving, too. The importance of educational algorithms have seen a rise in recent years as education increasingly grapples with the realities of the ‘age of big data’.
Ahead of this year‘s eLearning Africa conference in Ethiopia, the spotlight is on the country’s ambitious development goals. Alongside the Millenium Development Goals, in 2015, Ethiopia will reach the end of its five-year Growth and Transformation Plan (GTP). Central to the GTP’s ambitious aims, which include achieving and making Ethiopia a middle income country by 2020-23 (a target that would require an annual growth rate of 11.2% for the next 14 years), is improving the country’s technical and vocational training (TVET) capacities.
A recent World Bank Investment Climate Assessment has claimed that labour productivity in Ethiopia in important industries stands at less than half the average for Sub-Saharan Africa. As the government TVET strategy paper from 2008 outlined: ‘National TVET Strategy is an important element of the overall policy framework towards development and poverty reduction.’
It may be too early to predict how on-track Ethiopia is regarding these ambitious goals – although there has been good news recently regarding MDGs on infant mortality – it is possible to say that labour market issues persist.
A recent report into the state of the TVET field in Ethiopia has commented on the mechanisms of allocation in Ethiopia’s largely command-driven system. According to the authors Pramila Krishnan and Irina Shaorshadze, this system has led to inefficiency – often students applying for positions at TVET institutions are placed in positions radically different from their specialisation of choice. While researching their article, the authors found a group of graduates in major garment manufacturing plants who had applied to study engineering survey, and point to an apparently thriving informal placement exchange economy visible in the streets of Addis Ababa.
Could effective use of algorithms provide the answer? In their paper, Krishnan and Shaorshadze have pointed to a mechanism introduced in Boston for public school placement as a potential model for Ethiopian TVET placement. The mechanism, introduced in the 2005-6 school year, is known as a ‘student-optimal stable mechanism’ and uses a deferred-acceptance algorithm, replacing the old mechanism which allocated students on the basis of first and second choices. The authors claim that this mechanism is inherently inefficient and inequitable – in creating a hypothetical set of matching allocations that are superior and encouraging individuals to strategise by misrepresenting their choices. The new system is, on the contrary, strategy-proof, relying on the priorities of the children at each school and encouraging truth telling by incentivising the revealing of true preferences.
The authors note that the current TVET system runs broadly on the same framework of the old Boston system. Although the introduction of anything resembling the new Boston system to the Ethiopian TVET market would involve significant expansion and rethinking, it is certainly a prospect worth considering.
More broadly, algorithms seem to have a bright future in education and development. Massively scalable and free of cognitive and social bias, they are seen by many as representing a way pushing through traditional barriers to development. Just how much of a role they will play in development in the coming years is yet to be seen, but the evidence suggests it could be a vital one.