Team Brown HCRI
As robots increasingly take part in important areas of society
such as medicine, social care, education, or disaster response,
we must ensure that they follow the social and moral norms
of the communities they are part of. Currently, however,
robots follow only basic instructions without any conception
of social and moral norms. This, then, is the grand challenge
that the Brown HCRI team poses: to teach robots social and
moral norms. The team has initiated an interdisciplinary
research program that aims to meet this grand challenge in
three phases. In the identification phase, the team is developing experimental research methods and algorithms to
identify human norms for a subset of contexts and communities (for example, senior care, medical assistance, education).
Next, in the implementation phase, the team is building
computational networks of norms that have been identified
for the specific contexts. These networks must be flexible
enough to learn subtle context variations and to add or
update norms when receiving feedback from trusted sources.
Such feedback will come not only from people who interact
with the system, but also from crowdsourced observers who
are members of the relevant communities. Finally, in the
evaluation phase, the team will be installing these networks
in robots and evaluating their social acceptability in rigorous
human-robot interaction studies. Some of these studies will
take place in virtual and augmented-reality environments
that enable immersive experiences but also permit experimental control over critical causal variables, such as the
robot’s appearance or the transparency of its norm competence.
For more information, see hcri.brown.edu. The team contact is Bertram Malle ( email@example.com).
heavily developed market sector. Drug companies
have little incentive to develop methods for intelligently personalizing prescriptions since the intelligent agent may select the drugs of a competitor.
aifred health also excels in the systematic way they
are pursuing interdisciplinary research. In addition to
developing predictive models, they are developing
ethical frameworks to evaluate the performance of
Teams concerned with life well-being are attempting
to solve quality-of-life issues, including AI designs for
the hearing and vision impaired (three of four
advancing), personal life management (six of eleven
advancing), independent living assistance for the
elderly or infirm (one in five advancing), and one
team working to produce an online safety agent
(advancing). Several successful teams from these
groups are finding ways of promoting everyday well-
ness by extending the reach of clinical professionals
beyond the doctor’s office. The first-prize milestone
winner, Amiko AI, developed a model and sensors to
support the continuous monitoring of asthma treat-
ments. Amiko AI could easily be categorized a health
team, but their focus on facilitating the doctor and
patient relationship expands the boundaries of the
medical profession into the promotion of wellness.
Teams working on environmental problems are
developing solutions within the subcategories of
agriculture (one in four advancing), recycling (one of
one advancing), species abundance (one of one
advancing), water quality (zero of one advancing),
and pollution mitigation (one of one advancing).
WikiNet served as the pollution mitigation team and
received a nomination for a milestone award for their
work with the large unstructured corpus of environmental remediation documents to build a system
that can recommend best practices on future remediation efforts.
In this age of antibiotics, there is still an ongoing effort to discover new drugs to combat ill-nesses for which there is no known cure. In
addition, there is a need to discover replacements for existing drugs for pathogens that
have become resistant. Although multidrug
resistance in pathogens is growing fast, the
development of new drugs to treat bacterial
infections has reached its lowest point since
the beginning of the antibiotic era. The existing process for creating new drugs is slow, inefficient, and costly. DeepDrug is developing
eSynth, a drug design software that generalizes
from existing drug trial datasets to create an
improved method for identifying drug compounds.
eSynth can automatically synthesize targeted drug molecules, filter candidates based on
chemical criteria (such as being an antibiotic or
toxicity), analyze 3D image models of the
pathogen for possible drug repurposing, automate clinical testing for side effects, and predict the candidates most likely to succeed.
Recent progress includes design, training, and
testing of several AI filters and engines that
have shown promising results.
The team contact is Supratik Mukhopadhyay