IEEE Systems, Man and Cybernetics Magazine - July 2019 - 57

Lateral Position (m)

The second use case is in the
for the attribution of dependability
context of assistive robotics where
to an autonomous machine. FinalFaith in an
robots are required to provide perly, the development of trust is built
autonomous robot
sonal assistance to people with
on developing faith that the robot
specific needs. The capability of
will continue to perform the task in
will be rooted in
such robots to understand the
the future. Faith in an autonomous
predictability,
intentions and goals of the humans
robot will be rooted in predictabilithey are interacting with is of paraty, dependability, and extended
dependability, and
mount importance. Otherwise, if
experience with the robot.
extended experience
these robots fail in accurately
In addition, awareness of how
understanding and predicting the
autonomous robots can make our
with the robot.
intentions of the humans around
lives easier and enhance our qualthem, they could risk losing their
ity of life may be the first step
trust forever. Based on the aforeto achieving trust. The broader
mentioned reasoning and examples, we would strongly
society needs to be assured that these autonomous sysargue that creating intention-aware autonomous systems
tems are not competitors but allies.
is key to the upcoming integration of autonomous systems
into society.
Intention-Aware Trusted
Autonomous Systems
One of the critical factors that has been shown to strengthControl Is a Necessary Condition
en the human trust of autonomous systems is the ability of
for Gaining Trust When a Machine Is Involved
such systems to understand and comprehend human intenThe trust between human and machine is fundamentally
tions [16]. Human intentions, by their very nature, are a
different from the trust among humans. Often, the trust
complex property. However, some observable cues can be a
between human and machine is only about a very specifreliable indicator of human intentions. For example, the
ic aspect of the machine's functionality and is not generbody language of a pedestrian (shown in Figure 5) when
al in nature. In this article, I only discuss this specific
planning to cross the road [17] or the eye gaze of human
type of trust. I argue that having control over a machine
drivers when preparing to change lanes
[18] is very predictive of his or her intentions. While we use similar intention predictions in our daily lives, we have not
formalized the analysis method for them.
I argue that augmenting autonomous
systems with models of human intentions
and behaviors would open new avenues for
their application and certainly would make
them more empathetic trustworthy entities.
4
One of the use cases for the augmentation
3.5
IMM-CV/CA
technique is in the context of self-driving
3
Proposed
vehicles. Currently, many nonverbal com2.5
Ground Truth
munications occur among motorists in
2
1.5
human-driven vehicles and between them
1
and vulnerable drivers around them.
0.5
Exa mples of these communications
0
include the eye contact between motorists
-0.5
at intersections or roundabouts and a nod
-1
from a motorist to encourage a pedestrian
-1.5
to cross the road. Having a self-driving
-2
vehicle that can intuitively understand
-2.5
these kinds of intentions and implicit cues
10
0
-10
-20
-30
-40
-50
will increase the trust between society
TTE (Frames)
and autonomous systems. Furthermore, if
these self-driving vehicles can communiFigure 5. A sample prediction of the intended crossing action and
cate their understanding back to humans trajectory of a pedestrian using the framework proposed in [17]. TTE:
around them, that will cement the trust time to event; IMM: interacting multiple mode; CV: constant velocity;
building process.
CA: constant acceleration.
Ju ly 2019

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE

IEEE Systems, Man and Cybernetics Magazine - July 2019

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - July 2019