how to bid over the day, given a target specified by
the marketer and a budget. The problem involves
selecting the bid per impression, identifying the total
number of bids, and spreading the bids over the day.
The latter is pacing. Conitzer has established a pacing
equilibrium bidding strategy. The work is useful for
conducting auctions for several marketing actions
and is important in focusing attention on the problem of resource allocation in marketing. Given the
problem of endogeneity, another relevant task is
finding exogenous variation in data. One presentation presented a method for doing so.
Nair discussed the approach that marketing science takes to research, with the goal of informing
decision-making for business problems. At the start,
a model of consumer behavior is built, informed
both by economics and psychology. Aggregated
across consumers a demand function is obtained,
which has as arguments marketing actions. The
actions are optimized for appropriate business objectives. The approach is driven by theory, drawing
from received wisdom in consumer psychology and
economics. The need for prediction and scalability is
addressed by AI and ML, while the need to bring in
theory-driven models is met by MS. A common platform requires building models from the ground up,
rather than setting models up to learn from data.
Recognizing the process by which data is generated
becomes important for answering questions of causation. Rounding out the theme of understanding
consumer behavior for modeling were methods for
extracting brand perceptions from online data.
Weinstein presented the industry perspective on
AI’s future in addressing real-life marketing problems, including the broad area of advertising and
promotions. Weinstein’s deep experience across
industries, his insider perspective, and his insights
regarding the distinctions between first-party, sec-ond-party, and third-party data, and their different
uses when combined, provide new grounds for
research for both AI and MS communities. With new
research opportunities arises the need for new thinking about methods and models, much of which can
benefit from cross-pollination across AI and MS.
Paper presentations provided an understanding of
consumer behavior and response to adverts, useful
input for marketers in making advertising decisions.
The speakers, presenters, and participants agreed
that this workshop was a valuable experience in
bringing together academic and industry researchers
from both AI and MS. Attendees agreed that more
workshops should be organized to build on this
event, because the confluence of research is critical
for AI and MS to advance both research and practice
in this digital economy.
Hung Bui, Pradeep Chintagunta, S. Muthukrish-
nan, Tuomas Sandholm, Atanu R Sinha, and Geor-
gios Theocharous served as cochairs of this work-
shop. This report was written by Atanu Sinha and
Georgios Theocharous. The workshop papers were
published in the AAAI digital library.
Artificial Intelligence Applied to
Assistive Technologies and Smart
Recent progress in AI is reshaping the way we con-
ceive of the world. With the proliferation of sensors
and the lowering cost of smart devices, new subfields
of AI have emerged, such as assistive technology and
smart environments. A smart environment is a phys-
ical space, perhaps a living facility, equipped with
sensors, actuators, and AI capabilities, which togeth-
er provide smart services to the occupant. That kind
of environment has the potential to enhance quality
of life by providing assistance in the activities of dai-
ly living, rendering the environment itself a form of
assistive technology. Environments such as these are
particularly interesting for the support they can pro-
vide elders and impaired persons, improving their
autonomy and reducing the need for caregivers. The
market for assistive technologies is rapidly growing,
reaching $60.5 billion in 2018 in the US alone, with
an expected growth of almost 6 percent annually in
the next decade. Given this expansion, it has become
a major sector in research and development.
Despite the growing interest in these technologies,
however, they have not yet been widely adopted.
Indeed, the impairments and particularities of users
are so diverse that implementing solutions for well-being represents one of the major challenges of universal design. The goal of this workshop was to investigate new solutions to scientific problems in various
topics related to artificial intelligence as applied in
the domain of assistive technology and smart environments for persons with special needs.
The 2018 AAAI workshop brought together academic and industrial researchers from several subfields of AI. One main theme of the papers presented
at the workshop was learning methods used to personalize the assistance to a specific user. In fact, this
theme represents one major challenge in providing
good assistance: assistive systems need to learn precisely the profile of each of its users in order to be
able to help them effectively. Several papers presented learning approaches with validations and experiments on real data.
Another major theme was the development and
testing of new assistive devices. The papers on this
theme can be divided into two categories. The first
concerned new assistive devices to help populations
with physical impairments, such as smart arms for
wheelchairs. The second focused on devices for persons with cognitive deficits.
The workshop participants discussed the ways in
which assistive technologies can help create a better
future for populations with special needs, touching
too on how these technologies can benefit from