Anita began by asking Tarek about his broad research agenda, which she characterized as the application of digital technologies and particularly of AI in settings of extreme poverty and humanitarian emergency.    What do you see as the potential of digital tech in improving the lives of the poor?

   Tarek explained that, as he was finishing his PhD, the usage of mobile phones in low-income countries had reached 90%, which was up from 10% over the prior 15-year period.   The mobile phone had become a central piece of technology for delivering a whole range of services above voice-to-voice calling.   These services included digital payments in fragile states, which are defined by a basket of challenges that include poverty, exposure to risk, and insufficient capacity to respond to risk.   The technology created an opportunity to overcome a gap between the provision of humanitarian aid and its uptake by those at which it was targeted, which in this case was low-income Afghani women.

Anita asked Tarek about the specifics of the study, which is forthcoming in Management Science.  What did you find in your study?  Did the distribution of aid through mobile payments work?   How did recipients of the aid use the money?

   What Tarek and his co-authors found was that uptake of humanitarian aid through digital payments on mobile devices increased from 40% to more than 99% once the targeted recipients were trained in how to access and use the payments.  This paper is the culmination of a decade of study in Afghanistan that led to multiple papers reporting on interventions through both the the private and public sector.   Everyone was skeptical that digital distribution of payments could work.  What we found is that they don't work as well -- unless the gap between the provision of aid and the uptake of the provided aid is bridged.  In this study, we paid the recipients a test payment, and then called them to let them know it was available.   Immediately, recipients would take this payment and cash it in with a local merchant.  This occurred more than 99% of the time.   A simple human-centered design of the process led everything to work seamlessly.  Recipients knew what it meant to be paid, and they knew how to get value out of it.   The local policymakers had been deeply pessimistic, but ultimately became convinced that this could work.

Do you see this way of distributing aid as scalable to other settings?  What are the greatest challenges to its implementation?

   Tarek explained that he finds great potential in human-centered design.  It's important to look directly at the user experience and provide support to overcome that very last gap on implementation.

You have also studied how violent conflict near mines in Africa could potentially be mitigated by curbing industrial mining in places where artisanal mining is feasible.   What is the difference between industrial and artisanal mining, and why is industrial mining so much more deeply associated with conflict than artisanal mining?

    Tarek explained that he wouldn't make the claim that artisanal mining gets you better outcomes. He stepped back to explain that mining is quite important economically on the continent of Africa.  It makes up about 2/3rds of exports and about a quarter of GDP.   There's a large literature on what's called the "resource curse," which is about the political, social, and culture problems that arise when your economy is so dependent on the extraction of non-renewables.  In particular, conflict often arises over the control of these resources.  The dominant assumption is that much of this conflict involves armed rebel gangs, etc.   But that is not what we find.   We do three things in this paper.  First, we do a number of qualitative case studies in the Congo and Zimbabwe and other places.  These studies show that artisanal miners may work with hand tools at the fringe of industrial sites to get access to slag, tunnels, etc., and that conflict between and among them can arise as this occurs.  Second, we use machine learning to predict the geolocations in which this will occur -- often using satellite data to discern electromagnetic characteristics, etc.   Third, this allows us to create an indicator on where this might occur, which we can interact with shocks to prices.   It turns out that predicts violence even better.  So it's price shocks together with vulnerability to artisanal/industrial action that is tied to conflict.  Specifically, we find that about 1/3rd to 1/2 of the observed violence is happening in places with both artisanal and industrial mining because they are in tension with one another.  And the state may be complicit in the sense that it may have been brought in by industry to clear the site of the artisans.  The state may propagate the conflict.

The implication is that good regulatory oversight -- i.e., labor standards, human rights standards, concerns about these violent actors, know-your-customer policy, anti-money-laundering work, etc., -- can actually ensure that there's a bargain to be struck that doesn't lead to violence. That's a completely different mindset than thinking about the conflict as arising from rebels.

Anita pointed to a paper that Tarek wrote on how AI can help address predictable disasters.  There's a link in the chat given that we ran out of time to cover this comprehensively.

    Tarek explained that, with natural disasters, you see all of these immediate applications, decision makers can better assess risk because we got high quality data that helps us predict earthquakes.   That means that you can use machine learning to help target and deliver higher quality responses.  You can also get ahead of disasters through design, anticipatory financing, and  greater efficiencies in the logistics of disaster prevention.   There are also a range of issues around the ethics of AI that we try to address.   There are some rich challenges there. We wrote this about 4 years ago, and I think the progress of the technology is amazing.  But we need to be very sophisticated about how we deploy it.

What makes you optimistic?

   You can't control what happens to you, but you can control how you respond.  Our ingenuity in doing that effectively makes my optimistic.