bulk of my current
current research falls into several broad categories:
work aims to understand and overcome the Digital Divide, including
leapfrogging technology and policy designs.
- Better metrics
are important for understanding the digital divide. For
national data are misleading at best (e.g., official International Telecommunications Union [ITU] statistics),
and rural areas are the most underserved. In addition, most
composite measures focus on Literacy, wealth, and, of course,
penetration. I have proposed measures covering Awareness, Availability, Accessibility, and Affordability.
- The implications
of the digital divide
(network exclusion) can be far worse than previously explained by
network effect externality models, in part because of ICT's broad
reach, theoretical capabilities, and rapid changes (e.g., Web 2.0).
Using a new (proposed) exclusion centric framing of examining
disparity, Working with Ernest Wilson, I've shown that the costs
of exclusion can become exponential regardless of underlying
is often missing in much of the world, with lower speeds (including in
the upstream direction, caps on usage (I'm talking 0.5 GB/month),
port/application restrictions, or other restrictions on usage.
While network management has to be balanced with net
the far bigger challenge in much of the world is broadband
availability. Given mobiles are the norm for last-mile
this diminishes copper, necessary for DSL or cable. The "3rd
wire" into the home (electric) has theoretical potential, but I've
shown it is too little, too late in most cases (broadband
over powerline). Cost
of connectivity is a major problem, with limited backhaul
(uplinking) capabilities, and high costs. Some 17 policy
and regulatory add-ons can cost *multiple* times what the
technology could or should cost. This is one reason
connectivity in developing regions can cost vastly more in absolute
sense, not merely relative sense.
nations have the opportunity to leapfrog,
given their limited deployment of infrastructure to date.
are widespread, but they have limited capacity, (expensive) third generation (3G)
notwithstanding. Innovative technology (optical fibers plus
fixed/portable wireless) combined with open access models
can dramaticly lower the costs of connectivity. In the FiberAfrica project,
I've show a design whereby the majority of Africans can avail
nearby broadband for only about $1/capita investment. My work
also been focused on more effective regulation, which is a bigger
bottleneck than lack of capital per se. Developing regions
especially struggle with issues of public vs. private, part of the
political economy of reform and deregulation. There is, for example, a
significant amount of optical fiber in
developing regions, often owned by the electricity company.
challenge is who gets to use it, at what price, how, where, etc.
ICT for (Sustainable) Development,
has gained a lot of disability, including by governments and heads of
state. However, many efforts are simplistic, and lack proper
metrics for analysis of viability. In addition, there is
insufficient interaction and integration of the various domains
involved in a challenge, not least of which is the intersection of ICT
and development specialists. The field is littered with
studies, high visibility but limited scalability pilots, and solutions
that only make an incremental difference or that reinvent the wheel
I have, working with others, been
leading academic efforts to make ICT for Development a more rigorous,
scholarly, and impactfull research field. Based on several workshops
I organized for the UN, World Bank, and NSF, a major takeaway was the
importance and challenges of integrating development professions and
local stakeholders into ICTD projects, starting with the design and
goals phase all the way to the metrics and evaluation phase.
what extent is
ICT a means vs. and end? What are its limits, let alone costs vs.
benefits? The farmer may ultimately need
fertilizer, water, etc., but information can help optimize her inputs.
As an example pointed out by a participant at our first
Workshop, given nearly all
top 10 risk factors for burdens of disease in developing regions are
related to diet, sanitation, sex, indoor cooking, etc., educating and
empowering the mother becomes a primary option for
health in developing regions. I am actively working on these
questions, in part focusing on specific development domains.
specific research projects include:
use of speech recognition
on (mobile) phones for semi-literate healthcare workers (Healthline).
Roni Rosenfeld, Jahazeb Sherwani, and I hypothesize there is
a sweet spot between the literate (who
alternatives) and the completely illiterate (who may not have the
demand) for such technologies.
Hussain and I have shown
is an extremely cost-effective mechanism for providing information to
underserved communities. The capital costs are very modest,
and two-way communications are possible via dial-in by mobiles.
The challenges are often regulatory, where regulators have
traditionally regulated radios in a controlling manner or treated them
as commercial cash-cows.
parallels to other policy problems, Eswaran Subrahmanian and I have postulated
that ICTD might be a Wicked
Problem - one that is so severely over-constrained (or
under-constrained) that traditional engineering optimizations and
solutions will fail. One doesn't know which constraints to
relax (or add); the only solutions possible are iterative, and the
problem formulation is the key.
the enormous amounts of data that are and could be created, new
techniques for not only analyzing the data but developing policies may
be necessary, viz., machine
learning. Development agencies like the World
Bank extensively use tools like regressions, that apply linear (or
transformed linear) fits to data. What if the real world
doesn't look like this, and is better examined through regression trees
or neural nets? E.g., why is Brazil lumped together into S.
America for comparative analysis, when it has more in common with S.
Africa than Uruguay (at least per water data that EPP graduate student
Sean Green and we have examined)?
computer games be a
useful tool for teaching students, especially differently
abled children? With this goal, Robotics Institute students Ling Xu and
Vinitra Varadharajan created DeSIGN: An
Intelligent Tutor for American Sign Language.
- TechBridgeWorld is more
than a project, it is CMU university-wide initiative on bridging global
development through technology and partnerships. In addition
to courses, research, and outreach, a major component is student-driven
projects. Graduate students in many departments at CMU can
undertake research on topics of technology and development have have
this count towards their degree, a very unique system (V-Unit).
There are also opportunities for undergraduate and graduate
students to spend their summer doing ICTD consulting, implementation,
and design (TCinGC).
networks operate much as they did almost a century ago - the power
flows where it's requested based on uncontrolled physics.
newer technologies such as Flexible AC Transmission Systems (FACTS)
allow it to be controlled, we don't even have granular measurements of
power today. Smart metering is the idea that every
can be measured with arbitrary granularity, down to every consumer.
By moving beyond Time of Use pricing for electricity to
pricing, consumers can adapt their consumption to save money and also
make the grid more robust.
how the drivers
for smart metering are different in developing than
developed, focusing on peak management (avoiding all-too-common
blackouts) and theft reduction instead of labor saving per se.
key insights and questions:
ingredients for a cost-effective solution (down to households) exists,
but integrated solutions are not yet "off the shelf". The
materials for such systems is low - design and standards, in particular
may be key.
5% off the peak load
can reduce generation costs by about 25%; more for regions experiencing
the bulk of
savings can come from only a minority of consumers, what designs are
appropriate for the population at large? Are voluntary systems fair and
- A remote connect/disconnect switch
is important not merely for operation needs, but to allow emergency
load control in a fair, granular level. Today,
load shedding is crude, at a feeder level. In the future,
smart, ICT-enabled system, all homes could receive a lifeline of, say,
a few amps. Traffic lights, schools, hospitals, etc. (or
willing to pay through their nose) could receive full power.
are the appropriate stakeholders and how are they to be regulated?
the utility build its
own telecommunications system? What reliability are they
seeking? If the goal is auditing/accounting, delays are no problem.
If the goal is control,
then can they rely on a third party? Cellular networks are nearly
ubiquitous, but carriers can make more money off SMS (text) messages
than discount bulk messages from utilities. There is also the
economic issue - at 5 minute readings, this is over 100,000 readings
per node per year.
is or isn't
included in "the network?" Is there a gateway system? Where
logical boundary for the utility? They don't necessarily want to get
into the home, but they want to enable
smart appliances, distributed (renewable) generation, plug-in vehicles,
etc. Consider fridges - there is no reason the defrost cycle
(essentially, a heater, consuming much more peak load than the
compressor), ever turn on during peak periods? Talking to
manufacturers, in volume, the incremental cost of a communications chip
per fridge might be as little as a few dollars.
are consumer rights and privacy
to be met? An experimental project at CMU on "extreme
(sampling many times per second) shows that complete appliance-level
identification is possible!
what extent are solutions scalable, upgradeable, and
secure? The worst-case scenario would be if a hacker can turn
on a stove. Security is, in
fact, a major reason for bi-directional communications instead of
merely metering (billing) uploads.
am on the Technology
for a leading California utility's Advanced Metering project, a project
that was judged by its peers as the best in the US. Through CSTEP, I have been invited by
the Govt. of India to jointly create an updated report on IT for the Power
Energy and Power for Developing
to but perhaps worse than the digital divide is the global energy
divide. It's worse because the costs of providing services is
much higher than for ICT (mobiles, anyone?), and also because there are
global challenges of availability and the environment.
of my work has
focused on India, a booming economy constrained by its lack of
electricity and other infrastructure. There are more people
lacking electricity in S. Asia than in Africa!
of "India and China" as being to blame for high oil prices, food price
spikes, etc. There is no such thing as "India and China."
The latter has added electric capacity at over 100,000
while India is well below an order of magnitude that growth. China now
provides electricity to some 97+% of households, while in India the
majority of rural homes lack electricity. India has been
to grow energy services even without added constraints of carbon.
Asking it to comply with restrictions will likely fail.
China may comply due to air pollution concerns - India needs
technologies, finance, and assistance to meet its growth targets.
on specific topics:
Bharadwaj and I have been engaged in the analysis of biomass for power and energy
Conversion of biomass solids to electricity via gasification
somewhat proven technology (we set up a pilot plant in rural
Karnataka), but the issues are of scale and finance. Farmers
often get electricity for (nearly) free, and its really hard to compete
with free. One idea is to consider gasification for cooking
electricity - the former would displace the highly indoor polluting
solid cookstoves that lead to significant health issues in developing
to liquid fuels
is possible, but the challenges are more than economic
($140/barrel oil makes many alternatives competitive). Land
limited, and there can be detrimental impacts on food supply.
that just to create the tiny level of target blend for biodiesel
envisaged by the government of India would require cultivating
jatropha on an area bigger than the state of Tamil Nadu!
Dreams of marginal croplands, fallow lands, alongsides of
tracks may be wishful thinking because of the high supply-chain costs.
In addition, jatropha on dry land does grow, but the yield is
Similarly, for ethanol, until cellulosic technologies mature,
sugars will be market-constrained, and, more than that, in scale this
will require enormous land, water, etc. In contrast, just 4
SASOL style coal-to-liquid plants would produce enough diesel for a 20%
blend in the coming decades. Of course, India's coal is also
constrained, and there are environmental penalties, but this highlights
the need for integrated, life-cycle assessments of technologies, which we are working on during the coming years.
was never out of vogue in India (and is resurgant in the US), but
India's program has had a modest impact on total power (~2% of
I had shown that the 3-Phase
Nuclear Program as envisaged by India, reliant on
fast breeder reactors, would not grow capacity nearly as fast as hoped.
Instead, what may make sense, and currently being pursued, is
international collaboration for traditional nuclear power plants,
perhaps lightwater reactor design, using imported
fuel (India has limited domestic Uranium).
never dies - it
just needs updating! These are areas I don't spend as much time on
these topics as before, but try to keep updated or incorporated into
- Natural gas
is a clean,
versitile fuel, but its distribution is
regionally skewed. Unlike oil, there is limited global market given the
need for dedicated pipelines or special liquefaction facilities.
I calculated the costs and security measures
natural gas from West or Central Asia, overland
(i.e,. through Pakistan), showing it made economic
and strategic sense for both
countries. This work helped seed the proposed real-world
being negotiated today. I also did extensive briefings and
notes on this topic for government officials and decision-makers.
- It's easier to find out what's wrong with electricity pricing and regulation
than what's right. India's electricity two-part
tariff was flawed
because it gave the likelihood for additional profits to generators
without giving them any of the risk. I suggested some changes
the statutes, including pricing gas on actuals than nominal norms.
Electricity pricing is still contentious, but one needs a
of incentivizing producers and keeping costs low for consumers.
India operates mostly on costs-plus models, so getting it
is especially important. Of course, a market would only work
it's a true, competitive market!
- Electricity reform in developing regions hasn't gone as
(unbundling and privatization), as per the so-called World Bank model.
A major team effort headed by PESD, Stanford University,
this wasn't merely because countries didn't follow the textbook model
correctly - the model itself was flawed due to a poor understanding of
hybrid firms, ones combining public and private attributes.
chapter on the Political
Economy of Power Sector
India in The
Political Economy and Institutions of Power
Sector Reform in the Developing World,
I showed how India's reforms were surprisingly similar to the
experiences of other major developing countries, with some additional
twists due to the federal and state nature of electricity
is a theme that has
come up again and again, be it energy
security, or a
nebulous human security. Security is likely a relative term,
countries like India need not stress 100% indpendence as their goal
when it comes to energy (or food) security. It's too
costly, and hurts penetration and equitable distribution.