The Optimist’s
Guide To
The AI Governance
Challenge
How interactions
between humans and
machines can shape a
brighter future
TA B L E O F C O N T E N T S
2
Editor’s Note
4
A Magna Carta for Inclusivity and Fairness in
the Global AI Economy
12
Humans and AI: Rivals or Romance?
18
Why Machines Will Not Replace Us
24
The AI Governance of AI
31
AI and the Future of Lie Detection
37
Conclusion
1
Editor’s Note
Relevance
of
TITLE On
OFthe
THIS
CHAPTER
Behavioral Science to AI
SHALL GO HERE
2
On the Relevance of Behavioral
Science to AI
Even as we continue to develop a fuller understanding
of the mechanisms that underlay human judgment and
decision-making,
advancements
in
intelligent
machinery are forcing us to codify the processes by
which decisions are, and should be, made. Should your
self-driving Uber be allowed to break traffic
regulations in order to safely merge onto the highway?
To what extent are algorithms as prone to
discriminatory patterns of thinking as humans – and
how might a regulatory body make this
determination? More fundamentally, as more tasks are
delegated to intelligent machines, to what extent will
those of us who are not directly involved in the
development of these technologies be able to influence
these critical decisions? It is with these questions in
mind that we are pleased to have adapted the
following series of articles for publication at TDL.
– Andrew Lewis, Editor-in-Chief
3
Chapter One
A Magna
ta for
TITLE OF
THIS Car
CHAPTER
Inclusivit y and Fairness in
SHALL
GO
HERE
the Global AI Economy
4
A Magna Carta for Inclusivity
and Fairness in the Global AI
Economy
by Olaf Groth, Mark Nitzberg and Mark Esposito
We stand at a watershed moment for society’s vast,
unknown digital future.
A powerful technology,
artificial intelligence (AI), has emerged from its own
ashes, thanks largely to advances in neural networks
modeled loosely on the human brain. AI can find
patterns in massive unstructured data sets and
improve its own performance as more data becomes
available. It can identify objects quickly and accurately,
and make ever more and better recommendations —
improving
decision-making,
while
minimizing
interference from complicated, political humans. This
raises major questions about the degree of human
choice and inclusion for the decades to come. How will
humans, across all levels of power and income, be
engaged and represented? How will we govern this
brave new world of machine meritocracy?
5
Machine meritocracy
To find perspective on this questions, we must travel back
800 years: It was January, 1215, and King John of England,
having just returned from France, faced angry barons who
wished to end his unpopular vis et voluntas (“force and
will”) rule over the realm. In an effort to appease them, the
king and the Archbishop of Canterbury brought 25
rebellious barons together to negotiate a “Charter of
Liberties” that would enshrine a body of rights to serve as
a check on the king’s discretionary power. By June they
had an agreement that provided greater transparency and
representation in royal decision-making, limits on taxes
and feudal payments, and even some rights for serfs. The
famous “Magna Carta” was an imperfect document,
teeming with special-interest provisions, but today we
tend to regard the Carta as a watershed moment in
humanity’s advancement toward an equitable relationship
between power and those subject to it. It set the stage
eventually for the Enlightenment, the Renaissance, and
democracy.
Balance of power
It is that balance between the ever-increasing power of the
new potentate — the intelligent machine — and the power
of human beings that is at stake. Increasingly, our world
will be one in which machines create ever more value,
producing more of our everyday products. As this role
expands, and AI improves, human control over designs and
decisions will naturally decrease. Existing work and life
patterns will be forever changed. Our own creation is now
running circles around us, faster than we can count the
laps.
6
Machine decisions
This goes well beyond jobs and economics: in every area of
life machines are starting to make decisions for us without
our conscious involvement. Machines recognize our past
patterns and those of (allegedly) similar people across the
world. We receive news that shapes our opinions, outlooks,
and actions based on inclinations we’ve expressed in past
actions, or that are derived from the actions of others in
our bubbles. While driving our cars, we share our
behavioral patterns with automakers and insurance
companies so we can take advantage of navigation and
increasingly autonomous vehicle technology, which in
return provide us new conveniences and safer
transportation. We enjoy richer, customized entertainment
and video games, the makers of which know our
socioeconomic profiles, our movement patterns, and our
cognitive and visual preferences to determine pricing
sensitivity. As we continue to opt-in to more and more
conveniences, we choose to trust a machine to “get us
right.” The machine will get to know us in, perhaps, more
honest ways than we know ourselves — at least from a
strictly rational perspective. But the machine will not
readily account for cognitive disconnects between that
which we purport to be and that which we actually are.
Reliant on real data from our real actions, the machine
constrains us to what we have been, rather than what we
wish we were or what we hope to become.
7
Personal choice
Will the machine eliminate that personal choice? Will it do
away with life’s serendipity — planning and plotting our
lives so we meet people like us, thus depriving us of
encounters and friction that force us to evolve into
different, perhaps better human beings? There’s
tremendous potential in this: personal decisions are
inherently subjective, but many could be improved by
including more objective analyses. For instance, including
the carbon footprint for different modes of transportation
and integrating this with our schedules and pro-social
proclivities may lead us to make more eco-friendly
decisions; getting honest pointers on our more and less
desirable characteristics, as well as providing insight into
characteristics we consistently find appealing in others,
may improve our partner choices; curricula for large and
diverse student bodies could become more tailored to the
individual, based on the engine of information about what
has worked in the past for similar profiles.
Polarization
But might it also polarize societies by pushing us further
into bubbles of like-minded people, reinforcing our beliefs
and values without the random opportunity to check them,
defend them, and be forced to rethink them? AI might get
used for “digital social engineering” to create parallel
micro-societies. Imagine digital gerrymandering with
political operatives using AI to lure voters of certain
profiles into certain districts years ahead of elections, or
AirBnB micro-communities only renting to and from
certain socio-political, economic, or psychometric profiles.
Consider companies being able to hire in much more
surgically-targeted fashion, at once increasing their success
rates and also compromising their strategic optionality
with a narrower, less multi-facetted employee pool.
8
Who makes judgments?
A machine judges us on our expressed values — especially
those implicit in our commercial transactions — yet
overlooks other deeply held values that we have
suppressed or that are dormant at any given point in our
lives. An AI might not account for newly formed beliefs or
changes in what we value outside the readily-codified
realm. As a result, it might, for example, make decisions
about our safety that compromise the wellbeing of others
— doing so based on historical data of our judgments and
decisions, but resulting in actions we find objectionable in
the present moment. We are complex beings who regularly
make value trade-offs within the context of the situation at
hand, and sometimes those situations have little or no
codified precedent for an AI to process. Will the machine
respect our rights to free will and self-reinvention?
Discrimination and bias
Similarly, a machine might discriminate against people of
lesser health or standing in society because its algorithms
are based on pattern recognition and broad statistical
averages. Uber has already faced an outcry over racial
discrimination when its algorithms relied on zip codes to
identify the neighborhoods where riders were most likely
to originate. Will the AI favor survival of the fittest, the
most liked, or the most productive? Will it make those
decisions transparently? What will our recourse be?
9
Moreover,
a
programmer’s
personal
history,
predisposition, and unseen biases — or the motivations
and incentives of or from their employer — might
unwillingly influence the design of algorithms and sourcing
of data sets. Can we assume an AI will work with
objectivity all the time? Will companies develop AIs that
favor their customers, partners, executives, or
shareholders? Will, for instance, a healthcare-AI jointly
developed by technology firms, hospital corporations, and
insurance companies act in the patient’s best interest, or
will it prioritize a certain financial return?
We can’t put the genie back in the bottle, nor should we try
— the benefits will be transformative, leading us to new
frontiers in human growth and development. We stand at
the threshold of an evolutionary explosion unlike anything
in the last millennium. And like all explosions and
revolutions, it will be messy, murky, and fraught with
ethical pitfalls.
A new charter of rights
Therefore, we propose a Magna Carta for the Global AI
Economy — an inclusive, collectively developed, multistakeholder charter of rights that will guide our ongoing
development of artificial intelligence and lay the
groundwork for the future of human-machine co-existence
and continued, more inclusive, human growth. Whether in
an economic, social, or political context, we as a society
must start to identify rights, responsibilities, and
accountability guidelines for inclusiveness and fairness at
the intersections of AI and human life. Without it, we will
not establish enough trust in AI to capitalize on the
amazing opportunities it can and will afford us.
10
Moreover,
a
programmer’s
personal
history,
predisposition, and unseen biases — or the motivations
and incentives of or from their employer — might
unwillingly influence the design of algorithms and sourcing
of data sets. Can we assume an AI will work with
objectivity all the time? Will companies develop AIs that
favor their customers, partners, executives, or
shareholders? Will, for instance, a healthcare-AI jointly
developed by technology firms, hospital corporations, and
insurance companies act in the patient’s best interest, or
will it prioritize a certain financial return?
We can’t put the genie back in the bottle, nor should we try
— the benefits will be transformative, leading us to new
frontiers in human growth and development. We stand at
the threshold of an evolutionary explosion unlike anything
in the last millennium. And like all explosions and
revolutions, it will be messy, murky, and fraught with
ethical pitfalls.
A new charter of rights
Therefore, we propose a Magna Carta for the Global AI
Economy — an inclusive, collectively developed, multistakeholder charter of rights that will guide our ongoing
development of artificial intelligence and lay the
groundwork for the future of human-machine co-existence
and continued, more inclusive, human growth. Whether in
an economic, social, or political context, we as a society
must start to identify rights, responsibilities, and
accountability guidelines for inclusiveness and fairness at
the intersections of AI and human life. Without it, we will
not establish enough trust in AI to capitalize on the
amazing opportunities it can and will afford us.
11
Chapter Two
andCHAPTER
AI: Rivals
TITLEHumans
OF THIS
or Romance?
SHALL GO HERE
12
Humans and AI: Rivals or
Romance?
by Terence Tse, Mark Esposito and Danny Goh
We stand at a watershed moment for society’s vast,
unknown digital future.
A powerful technology,
artificial intelligence (AI), has emerged from its own
ashes, thanks largely to advances in neural networks
modeled loosely on the human brain. AI can find
patterns in massive unstructured data sets and
improve its own performance as more data becomes
available. It can identify objects quickly and accurately,
and make ever more and better recommendations —
improving
decision-making,
while
minimizing
interference from complicated, political humans. This
raises major questions about the degree of human
choice and inclusion for the decades to come. How will
humans, across all levels of power and income, be
engaged and represented? How will we govern this
brave new world of machine meritocracy?
13
The (human) empire strikes back
It’s certain that we will hear more and more alarmist
accounts. However, we have seen it before – many times, in
fact. Back in 1963, it was J F Kennedy who said, “We have a
combination of older workers who have been thrown out
of work because of technology and younger people coming
in […] too many people are coming into the labor market
and too many machines are throwing people out” [4].
Going further back, when the first printed books with
illustrations started to appear in the 1470s in Germany,
wood engravers protested as they thought they would no
longer be needed [5].But this all begs one question: If
technological progress represents a comprehensive threat
to humans, then why do we still have jobs left? In fact,
many of us are still working, probably much harder than
before. The answer: machines and humans excel in
different activities. For instance, machines are frequently
no match for our human minds, senses and dexterity. For
example, even though Amazon’s warehouses are
automated, humans are still required to do the actual
shelving.
And this doesn’t only apply to physical jobs. The real story
behind today’s AI is that it cannot function without
humans in the loop. Google is thought to have 10,000
‘raters’ who look at YouTube videos or test new services.
Microsoft, on the other hand, has a crowdsourcing
platform called Universal Human Relevance System to
handle a great deal of small activities, including checking
the results of its search algorithms [6]. And this blend of AI
and humans, who follow through when the AI falls short, is
not going to disappear any time soon [7]. Indeed, the
demand for such on-demand human interventions is
expected to continue to grow. The ‘human cloud’ is set to
boom.
14
Closer together
The above illustrates a very important lesson – humans
will be needed. The key is how to integrate humans and
machines in various activities and how to steer AI towards
the creation of new economic interfaces, rather than
towards the mere replacement/displacement of existing
ones. At the moment, the probability of AI getting things
right is between 85 and 95 per cent. Humans, on the other
hand, generally score 60 to 70 per cent. On this basis alone,
we need only machines and not humans.
Yet, in some highly data-driven industries such as financial
and legal services, there can be no error – any mistake can
result in huge financial costs in the form of economic
losses or expensive lawsuits. Machines by themselves are
not enough. Furthermore, AI can only run an algorithm
that is predefined and trained by a human, and so a margin
of error will always exist. When mistakes take place, AI will
not be able to fix them. Humans, by contrast, are able to
create solutions to problems. We believe that the best
solution is to use machines to run production up to the
level of 95 per cent accuracy, and supplement this with
human engineers to mitigate risks if not to strive to
improve accuracy.
Humans and machines will – and must – work together. As
business consultants, educators and policy advisors, we all
strongly believe that, ultimately, what really matters is how
to prepare people to work increasingly closely with
machines.
15
Closer together
The above illustrates a very important lesson – humans
will be needed. The key is how to integrate humans and
machines in various activities and how to steer AI towards
the creation of new economic interfaces, rather than
towards the mere replacement/displacement of existing
ones. At the moment, the probability of AI getting things
right is between 85 and 95 per cent. Humans, on the other
hand, generally score 60 to 70 per cent. On this basis alone,
we need only machines and not humans.
Yet, in some highly data-driven industries such as financial
and legal services, there can be no error – any mistake can
result in huge financial costs in the form of economic
losses or expensive lawsuits. Machines by themselves are
not enough. Furthermore, AI can only run an algorithm
that is predefined and trained by a human, and so a margin
of error will always exist. When mistakes take place, AI will
not be able to fix them. Humans, by contrast, are able to
create solutions to problems. We believe that the best
solution is to use machines to run production up to the
level of 95 per cent accuracy, and supplement this with
human engineers to mitigate risks if not to strive to
improve accuracy.
Humans and machines will – and must – work together. As
business consultants, educators and policy advisors, we all
strongly believe that, ultimately, what really matters is how
to prepare people to work increasingly closely with
machines.
16
References
[1] Frey, Carl Benedikt and Osborne, Michael. The Future of Employment: How
susceptible are jobs to computerisation? Oxford Martin School, 2013.
http://www.oxfordmartin.ox.ac.uk/publications/view/1314
[2] Chui, Michael, Manyika, James, and Miremadi, Mehdi. How Many of Your
Daily Tasks Could Be Automated?, Harvard Business Review, 14 December
2015. (https://hbr.org/2015/12/how-many-of-your-daily-tasks-could-beautomated)
[3] Tse, Terence and Esposito, Mark. Understanding How the Future Unfolds:
Using Drive to Harness the Power of Today’s Megatrends. Lion Crest, 2017.
[4] John F Kennedy interview by Walter Cronkite, 3 September 1963,
https://www.youtube.com/watch?v=RsplVYbB7b8
[5] The Economist. Artificial intelligence will create new kinds of work, 26
August 2017. https://www.economist.com/news/business/21727093humans-will-supply-digital-services-complement-ai-artificial-intelligence-willcreate-new
[6] Ibid.
[7] Gray, Mary L. and Suri, Siddharth. “The humans working behind the AI
curtain,”Harvard Business Review, 9 January 2017.
17
Chapter Three
Machines
Will
TITLE Why
OF THIS
CHAPTER
Not Replace Us
SHALL GO HERE
18
Why Machines Will Not
Replace Us
by Terence Tse, Mark Esposito and Danny Goh
Lately, we have received quite a number of requests
asking us to explain further why artificial intelligence
(AI) and robots are unlikely to put humans out of work
soon. It may be a contrarian position, but we are
definitely optimistic about the future, believing that
the displacement of labour won’t turn out to be as
gloomy as many are speculating. Despite the endless
talk on the threat of machines to human jobs, the truth
is that, while we have lost jobs in some areas, we have
gained them in others. For instance, the invention of
automatic teller machines (ATMs), introduced in the
1960s, ought to have eliminated the need for many
bank employees in the US. Yet, over time, the industry
has not just hired more staff, but job growth in the
sector is, in fact, doing better than the average [1].So,
why is this? The answer can actually be found in
Hollywood movies. In the 1957 film Desk Set, the
entire audience research department in a company is
about to be replaced by a giant calculator.
19
It is a relief to the staff, however, when they find out that
the machine makes errors, and so they get to keep their
jobs, learning to work alongside the calculator. Fast
forward to the 2016 film Hidden Figures. The human
‘computers’ at NASA are about to be replaced by the newly
introduced IBM mainframe. The heroine, Dorothy Vaughan,
decides to teach herself Fortran, a computer language, in
order to stay on top of it. She ends up leading a team to
ensure the technology performs according to plan.
Facts and not fantasies
These are not merely fantasies concocted by film studios.
Granted, realistically, many jobs, especially those involving
repetitive and routine actions, may succumb to automation
for good. But the movies above do encourage us not to
overrate computers and underrate humans. Delving
deeper into this, we believe there are several elements that
underpin this message.
Only humans can do non-standardised tasks. While
traditional assembly line workers are set to be replaced by
automation, hotel housekeeping staff are unlikely to face
the same prospect any time soon. Robots are good at
repetitive tasks but lousy at dealing with varied and
unique situations. Jobs like room service require flexibility,
object recognition, physical dexterity and fine motor
coordination; skills like these are – at the moment at least
– beyond the capabilities of machines, even for those
considered intelligent.
20
Machines make human skills more important. It is possible
to imagine an activity – such as a mission or producing
goods – to be made up of a series of interlocking steps, like
the links in a chain.
A variety of elements goes into these steps to increase the
value of the activity, such as labour and capital; brain and
physical power; exciting new ideas and boring repetition;
technical mastery and intuitive judgement; perspiration
and inspiration; adherence to rules; and the considered
use of discretion. But, for the overall activity to work as
expected, every one of the steps must be performed well,
just as each link in a chain must do its job for the chain to
be complete and useful. So, if we were to make one of these
steps or links in a chain more robust and reliable, the value
of improving the other links goes up [2]. In this sense,
automation does not necessarily make humans
superfluous.
Not in any fundamental way, at least; instead, it increases
the value of our skill sets. As AI and robots emerge, our
expertise, problem-solving, judgement and creativity are
more important than ever [3]. For example, a recent study
looks into a Californian tech startup. Despite the company
providing a technology-based service, it finds itself to be
growing so fast that, with the computing systems getting
larger and more complex, it is constantly drafting in more
humans to monitor, manage and interpret the data [4].
Here, the technologies are merely making the human skills
more valuable than before.
21
Social aspects matter. Perhaps one of the most telling
lessons learnt from underestimating the power of human
interactions can be found by looking at Massive Open
Online Courses (MOOCs). Until recently, it was widely
believed that the rise of digital teaching tools would make
human teachers less relevant, or even superfluous.
However, that was not found to be the case with MOOCs.
Instead, they have shown that human teachers can be
made more effective with the use of digital tools.
The rise of hybrid programmes, in which online tools are
combined with a physical presence, has only partially
reduced the number of face-to-face hours for teachers,
while freeing them up to be more involved with curriculum
design, video recording and assessment writing.
Ultimately, it is this combination of human interactions and
computers that champions [5].Human resistance is not
futile. Many of us have witnessed seemingly promising IT
projects end up in failure. But, very often, this is not the
result of technological shortcomings. Instead, it is the
human users that stand in the way. Unfamiliar interfaces,
additional data-entry work, disruptions to routines, and
the necessity to learn and understand the goals the newly
implemented system is trying to achieve, for instance,
often cause frustration.
The aftermath is that people can put up an enormous
amount of resistance to taking on novel technologies, no
matter how much the new systems would benefit them
and the company. Such an urge to reject new systems is
unlikely to change in the short term.
22
Closer together
There is simply no reason to think that AI and robots will
render us redundant. It is projected that, by 2025, there
will be 3.5 million manufacturing job openings in the US,
and yet 2 million of them will go unfilled because there will
not be enough skilled workers [6]. In conclusion, rather
than undermining humans, we are much better off
thinking hard about how to upskill ourselves and learn
how to work alongside machines, as we will inevitably
coexist – but it won’t be a case of us surrendering to them.
References
[1] Bessen, James. Learning by Doing: The Real Connection Between Innovation,
Wages, and Wealth. Yale University Press, 2015.
[2] This is called the O-ring theory or principle, which was put forward by
Michael Kremer in 1993. The name comes from the disaster of the space
shuttle Challenger in 1986, which was caused by the failure of a single O-ring.
In this case, an inexpensive and seemingly inconsequential rubber O-ring seal
in one of the booster rockets failed after hardening and cracking during the icy
Florida weather on the night before the launch.
[3] Autor, David. “Why Are There Still So Many Jobs? The History and Future of
Workplace Automation”, Journal of Economic Perspectives, 29(3), 3-30, 2015.
[4] Shestakofsky, Benjamin. “Working Algorithms: Software Automation and
the Future of Work”, Work and Occupations, 44(4), 2017.
[5] Tett, Gillian. “How robots are making humans indispensable”, Financial
Times, December 22, 2016. https://www.ft.com/content/da95cb2c-c6ca11e6-8f29-9445cac8966f
[6] Manufacturing Institute and Deloitte, Skills Gap in US Manufacturing, 2017.
https://www2.deloitte.com/us/en/pages/manufacturing/articles/skills-gapmanufacturing-survey-report.html
23
Chapter four
Governance
of
TITLEThe
OFAITHIS
CHAPTER
AI
SHALL GO HERE
24
The AI Governance of AI
by Josh Entsminger, Mark Esposito, Terence Tse and Danny Goh
Our lives are ruled by data. Not just because it informs
companies of what we want, but because it helps us to
remember and differentiate what we want, what we
need, and what we can ignore. All these decisions give
way to patterns, and patterns, when aggregated, give
us a picture of ourselves. A world where such patterns
follow us, or even are sent ahead of us — to
restaurants to let them know if we have allergies, to
retail stores to let them know our preferred clothing
size —is now so feasible that labeling it science fiction
would expose a lack of awareness more than a lack of
imagination.
The benefits of AI are making many of our choices
easier and more convenient, and in so doing,
tightening competition in the space of customer
choice. As this evolution happens, the question is less
to what extent AI is framing our choices, but rather,
how it is shaping them. In such a world, we need to
understand when our behavior is being shaped, and by
whom.
25
Clearly, most of us are quite comfortable living in a world
where our choices are shaped by AI. Indeed, we already
live in such a world: from search engines to smooth traffic
flow, many of our daily conveniences are built on the speed
provided by the backend. The question we need to ask
ourselves when considering AI, and its governance, is
whether we are comfortable living in a world where we do
not know if, and how, we are being influenced.
Behavioral Bias or Behavioral Cues
AI can do for our understanding of behavior what the
microscope did for biology.
We have already reached the point where software can
discover tendencies in our behavioral patterns we might
not consider, identifying traits that our friends and family
would not know. The infamous, but apocryphal, story of
the father who discovered his daughter was pregnant
when Target began sending advertisements for baby
suppliers (after detecting a shift in her spending) gives us a
sneak peek.[2]
Our lives are already ruled by probabilistic assumptions,
intended to drive behavior. Now we need to ask, and
answer honestly, how much of your life are you willing to
have shaped by algorithms you do not understand? More
importantly, who should be tasked with monitoring these
algorithms to flag when they have made a bad decision, or
an intentionally manipulative one?
26
As more companies use AI, and the complexity of its
insights continues to grow, we will be facing a gap above
the right to an understanding, or the right to be informed –
we will be facing a gap concerning when, and if, a violation
has occurred at all.
As our digital presence grows, and this presence is being
pulled by public directions for the future of e-governance
and private for how we engage with our interests –
meaningful governance will have to include an essential
first step, the right to know how our data is being used,
who has it, and when are they using it.
Another example of how behavior and technology are
interfacing at a faster than ever pace is through the
observation of what Chatbots have been shown to provide
us: the potential for emotional associations, which might
be used for manipulative purposes.[2] As developments in
natural language processing grow to combine with
advanced robotics, the potential of building that bond from
touch, warmth, comfort, also grows – particularly in a
world where we experience the largest endemic of
loneliness, driving the UK to literally appoint a minister for
loneliness.
As machine to machine data grows in the internet of
things, companies with preferential access will have more
and more insight into more and more minute aspects of
behavioral patterns we ourselves might not understand —
and with that comes a powerful ability to nudge behavior.
Good data is not just about volume, it is about veracity —
as IoT grows, we are handing firms everything they need to
know about us on a silver platter.
27
We can argue still that the issue is not the volume, the
issue is the asymmetry of analytic competency in
managing that volume — meaning asymmetries in
capturing value. In turn, this means some companies not
only understand you, but can predict your behaviour to the
point of knowing how to influence a particular choice most
effectively. In the age of big data, the best business is the
insight business.
Accountability: who is looking after us?
The first question concerning building accountability is
how to keep humans in the decision loop of processes
made more autonomous through AI. The next stage needs
to preserve accountability in the right to understanding —
to know why an algorithm made one decision instead of
another.
New proposals are already emerging on how to do this —
for example, when specific AI projects are proprietary
aspects of a firm’s competitiveness, we might be able to use
counterfactual systems to assess all possible choices faced
by an AI.[3] But systems that map decisions without
breaking the black box will not be able to provide the
rationale by which that algorithm made one decision
instead of another.
28
Yet the problem still goes deeper. The problem with
transparency models is the assumption that we will even
know what to look for — that we will know when there
needs to be a choice in opting out of a company’s use of
data. In the near future, we may not be able to understand
by ourselves when an AI is influencing us.
This leads us to a foundational issue: to govern AI, we may
need to use AI.
We will need AI not just to understand when we are being
influenced in overt ways, but to understand the new
emerging ways in which companies can leverage the
micro-understanding of our behavior. The capacity for
existing legal frameworks, existing political institutions,
and existing standards of accountability to understand,
predict, and catch the use of AI for manipulative purposes
is sorely lacking.
Algorithmic collusion is already a problem — with pricing
cartels giving way to pop-up pricing issues that can
disappear, without prior agreement, thus avoiding the
initial claims.[4] We can imagine a world where collusion
is organized not by the market, but by tracking the
behavior of distinct groups of individuals to organize
micro-pricing changes.
Naturally, questions emerge: who will govern the AI that
we use to watch AI? How will we know that collusion is not
emerging between the watchers and the watched? What
kind of transparency system will we need for a governing
AI to minimize the transparency demands for corporate
AI?
29
The future of AI governance will be decided in the margins
— what we need to pay attention to is less the shifting
structure of collusion and manipulation, but the conduct,
and the ability for competent AI to find the minimal
number of points of influence to shape decision making.
We need to have a conversation to make our assumptions
and beliefs about price fixing, about collusion, about
manipulation, painfully clear. In an age of AI, we cannot
afford to be vague.
References
[1] Piatetsky, Gregory. Did Target really predict a teens pregnancy? The Inside
Story May 7 2014 KD nuggets
[2] Yearsly, Liesl. We need to talk about the power of AI to manipulate humans.
June 5 2017. MIT Tech Review
[3] Mittelstadt, Brent. Wachter, Sandra. Could counterfactuals explain
algorithmic decisions without opening the black box? 15 January 2018. Oxford
Internet Institute Blog
[4] Algorithms and Collusion: Competition Policy in the Digital Age. OECD
2017
30
Chapter five
theCHAPTER
Future of
TITLEAIOFand
THIS
Lie Detection
SHALL GO HERE
31
AI and the Future of Lie
Detection
by Josh Entsminger, Mark Esposito, Terence Tse and Danny Goh
We live in a world now where we know how to lie.
With advances in AI, it is very likely that we will
soon live in a world where we know how to detect
truth. The potential scope of this technology is vast
— the question is how should we use it?
Some people are naturally good liars, and others are
naturally good lie detectors. For example, individuals
who fit the latter description can often sense lies
intuitively, observing fluctuations in pupil dilation,
blushing, and a variety of micro-expressions and body
movements that reveal what’s going on in someone
else’s head. This is because, for the vast majority of us
who are not trained deceivers, when we lie, or lie by
omission, our bodies tend to give us away.
For most of us, however, second guessing often
overtakes intuition about whether someone is lying.
Even if we are aware of the factors that may indicate a
lie, we are unable to simultaneously observe and
process them in real time — leaving us, ultimately, to
guess whether we are hearing the truth.
32
Now suppose we did not have to be good lie detectors,
because we would have data readily available to know if
someone was lying or not. Suppose that, with this data, we
could determine with near-certainty the veracity of
someone’s claims. We live in a world now where we know
how to lie. With advances in AI, it is very likely that we will
soon live in a world where we know how to detect truth.
The potential scope of this technology is vast — the
question is how should we use it?
The Future of AI Lie Detection
Imagine anyone could collect more than just wearable data
showing someone’s (or their own) heartbeat, but
continuous data on facial expressions from video footage,
too. Imagine you could use that data, with a bit of training,
to analyze conversations and interactions from your daily
life — replaying ones you found suspicious with a more
watchful gaze. Furthermore, those around you could do the
same: imagine a friend, or company, could use your past
data to reliably differentiate between your truths and
untruths, matters of import and things about which you
could not care less.
This means a whole new toolkit for investigators, for
advertisers, for the cautious, for the paranoid, for
vigilantes, for anyone with internet access. Each of us will
have to know and understand how to manage and navigate
this new data-driven public record of our responses.
33
The issue for the next years is not whether lying will be
erased — of course it will not —but rather, how these new
tools should be wielded in the pursuit of finding the truth.
Moreover, with a variety of potential ways of mis-reading
and misusing these technologies, in what contexts should
they be made available, or promoted?
The issue for the next years is not whether lying will be
erased — of course it will not —but rather, how these new
tools should be wielded in the pursuit of finding the truth.
Moreover, with a variety of potential ways of mis-reading and
misusing these technologies, in what contexts should they be
made available, or promoted?
The Truth About Knowing the Truth
Movies often quip about the desire to have a window into
someone else’s brain, to feel assured that what they say
describes what they feel, that what they feel describes
what they will do, and what they will do demonstrates
what everything means for them. Of course, we all know
the world is not so neat, and one might fall prey to
searching for advice online. What happens when such
advice is further entrenched in a wave of newly available,
but poorly understood, data?
34
What will happen, for example, when this new data is used
in the hiring process, with candidates weeded out by
software dedicated to assessing whether and about what
they’ve lied during an interview? What will happen when
the same process is used for school selection, jury
selection, and other varieties of interviews, or when the
results are passed along to potential employers. As the
number of such potential scenarios grows, the question we
have to ask is when is our heartbeat private information?
Is knowledge of our internal reactions itself private, simply
because until now only a small segment of perceptive
people could tell what was happening? Communities often
organize around the paths of least resistance, creating a
new divide between those who understand and can
navigate this new digital record, and those who cannot.
Imagine therapists actively recording cognitive dissonance,
news shows identifying in real time whether or not a guest
believes what they are saying, companies reframing
interviews with active facial analysis, quick border security
questioning. The expanding scope of sensors is pushing us
away from post-truth to an age of post-lying, or rather, an
end to our comfort with the ways in which we currently lie.
As with everything, the benefits will not be felt equally.
We might even be able to imagine the evolution of lie
detection moving towards brain-computer interfaces —
where one’s right to privacy must then be discussed in
light of when we can consider our thoughts private.
35
In court rooms, if we can reliably tell the difference
between reactions during a lie and during the truth, do
witnesses have a right to keep that information private?
Should all testimonials be given in absolute anonymity?
Researchers at the University of Maryland developed
DARE, the Deception Analysis and Reasoning Engine,
which they expect to be only a few years away from near
perfect deception identification.
How then should we think about the 5th amendment of the
US constitution, how should we approach the right to not
incriminate oneself? With the advent of these technologies,
perhaps the very nature of the courtroom should change.
Witnesses are not given a polygraph on the stand for good
reason: it’s unreliable — but there may be little stopping
someone with a portable analytics system to tell their
vitals or analyze a video feed from a distance, and publish
the results for the court of public opinion. How should our
past behavior be recorded and understood?
How we design nudges, how we design public spaces, how
we navigate social situations, job offers, personal
relationships, all depend on a balance of social convention
by which we allow ourselves — and others — to hide
information. Yet, what should we do with a technology that
promises to expose this hidden information? Is a world
based on full-truths preferable to the one we have now?
Will we have a chance to decide?
Advances in AI and the democratization of data science are
making the hypothetical problem of what kind of world we
prefer an all too real discussion, one we need to have soon.
Otherwise we’ll have no say in determining what lies
ahead.
36
Closing Remarks
OnOF
TheTHIS
Future
Discourse
TITLE
CHAPTER
Around AI & Behavior
SHALL GO HERE
37
On The Future Discourse
Around AI & Behavior
Much of the conversation around AI has been on the
impact it will have on our working lives. While this is
an important issue and certainly a very visible one
these days, in many ways it only scratches the surface
of the true impact AI is poised to have on humanity.
As AI becomes more and more commoditized and
becomes a bigger part of our physical and social
environment, we believe the conversation will begin to
shift from critical to constructive - from 'Will AI pose a
huge threat to human welfare?' to 'How can we make
sure the impact of AI on welfare is as sustained and
equitable as possible'. With this in mind, we believe
that it is crucial to start a discourse about the
intersection of AI and behavior science, in particular
the behavioral economics and social psychology of
human-computer interactions in a world run by
algorithms. We hope that this document has been a
first step in a direction toward this AI-humanism.
38
TITLE OF Authors
THIS CHAPTER
SHALL GO HERE
39
Authors
Dr. Olaf Groth
Olaf Groth, Ph.D. is co-author of “Solomon’s Code” and CEO of Cambrian.ai, a
network of advisers on global innovation and disruption trends, such as AI,
IOT, autonomous systems and the 4th Industrial Revolution for executives and
investors. He serves as Professor of Strategy, Innovation & Economics at Hult
International Business School, Visiting Scholar at UC Berkeley’s Roundtable on
the International Economy, and the Global Expert Network member at the
World Economic Forum.
Dr. Mark Nitzberg
Mark Nitzberg, Ph.D. is co-author of “Solomon’s Code” and Executive Director
of the Center for Human-Compatible AI at the University of California at
Berkeley. He also serves as Principal & Chief Scientist at Cambrian.ai, as well as
advisor to a number of startups, leveraging his combined experience as a
globally networked computer scientist and serial social entrepreneur.
Mark Esposito
Mark Esposito is a member of the Teaching Faculty at the Harvard University's
Division of Continuing, a Professor of business and economics, with
appointments at Grenoble Ecole de Management and Hult International
Business School. He is an appointed Research Fellow in the Circular Economy
Center, at the University of Cambridge's Judge Busines School. He is also a
Fellow for the Mohammed Bin Rashid School of Government in Dubai. At
Harvard, Mark teaches Systems Thinking and Complexity, Economic Strategy
and Business, Government & Society for the Extension and Summer Schools
and serves as Institutes Council Co-Leader, at the Microeconomics of
Competitiveness program (MOC) developed at the Institute of Strategy and
Competitiveness, at Harvard Business School. He is Founder & Director of the
Lab-Center for Competitiveness, a think tank affiliated with the MOC network
of Prof. Michael Porter at Harvard Business School and Head of the Political
Economy and Sustainable Competitiveness Initiative. He researches the
"Circular Economy" inside out and his work on the topic has appeared on top
outlets such as The Guardian, World Economic Forum, Harvard Business
Review, California Management Review, among others. He is the co-founder of
the concepts of "Fast Expanding Markets" and "DRIVE", which represent new
lenses of growth detection at the macro, meso and micro levels of the economy.
He is also an active entrepreneur and co-founded Nexus FrontierTech, an
Artificial Intelligence Studio, aimed at providing AI solutions to a large
portfolio of clients. He was named one of the emerging tomorrow's thought
leaders most likely to reinvent capitalism by Thinkers50, the world’s premier
ranking of management thinkers and inducted into the "Radar" of the 30 most
influential thinkers, on the rise.
40
Authors
Terence Tse
Terence is a co-founder & managing director of Nexus Frontier Tech: An AI
Studio. He is also an Associate Professor of Finance at the London campus of
ESCP Europe Business School. Terence is the co-author of the bestseller
Understanding How the Future Unfolds: Using DRIVE to Harness the Power of
Today’s Megatrends. He also wrote Corporate Finance: The Basics. In addition
to providing consulting to the EU and UN, Terence regularly provides
commentaries on the latest current affairs and market developments in the
Financial Times, the Guardian and the Economist, as well as through CNBC and
the World Economic Forum. He has also appeared on radio and television
shows and delivered speeches at the UN, International Monetary Fund and
International Trade Centre. Invited by the Government of Latvia, he was a
keynote speaker at a Heads of Government Meeting, alongside the Premier of
China and Prime Minister of Latvia. Terence has also been a keynote speaker at
corporate events in India, Norway, Qatar, Russia and the UK. Previously,
Terence worked in mergers and acquisitions at Schroders, Citibank and Lazard
Brothers in Montréal and New York. He also worked in London as a consultant
at EY, focusing on UK financial services. He obtained his PhD from the Judge
Business School at the University of Cambridge.
Danny Goh
Danny is a serial entrepreneur and an early stage investor. He is the partner
and Commercial Director of Nexus Frontier Tech, an AI advisory business with
presence in London, Geneva, Boston and Tokyo to assist CEO and board
members of different organisations to build innovative businesses taking full
advantage of artificial intelligence technology.
Danny has also co-founded
Innovatube, a technology group that operates a R&D lab in software and AI
developments, invests in early stage start-ups with 20+ portfolios, and acts as
an incubator to foster the local start-up community in South East Asia.
Innovatube labs have a team of researches and engineers to develop cutting
edge technology to help start-ups and enterprises bolster their operation
capabilities. Danny currently serves as an Entrepreneurship Expert with the
Entrepreneurship centre at the Said Business School, University of Oxford and
he is an advisor and judge to several technology start-ups and accelerators
including Startupbootcamp IoT London. Danny has lived in four different
continents in the last 20 years in Sydney, Kuala Lumpur, Boston and London,
and constantly finds himself travelling.
41
Authors
Josh Ensminger
Josh Entsminger is an applied researcher at Nexus Frontier Tech. He
additionally serves as a senior fellow at Ecole Des Ponts Business School’s
Center for Policy and Competitiveness, a research associate at IE business
school’s social innovation initiative, and a research contributor to the world
economic forum’s future of production initiative.
42
Thank you for reading!
Want to work together on a problem where
behavior matters?
Email Us: info@thedecisionlab.com
Visit Us: www.thedecisionlab.com
43