IEEE Technology and Society Magazine - December 2019 - 24

weapons release occurs via a pre-
scribed set of rules of engagement.
For these targets, the most meaning-
ful human control likely comes at
the point of final target approval.
Indeed, the job of the pilot for
these missions is primarily to get
the weapon to the point in the sky
where it can be released. The pilot
can execute some judgment and
call off a mission if various param-
eters are not met, such as civilians
in the area. Yet this is problematic
because workload is extremely high
for a pilot in any weapons release
scenario. In addition to making
sure the plane is in the correct loca-
tion, the pilot needs to consider any
enemy activity such as surface-to-
air missiles, which is very stressful.
This combination of high workload
and stress also raises the likelihood
of human error. So is this kind of
operation truly meaningful human
control? Because of these issues
and improvements in technology,
this job can be and has been suc-
cessfully replaced by a smart mis-
sile in many instances, suggesting
that role allocation has shifted from
the human to the weapon because
of advancements in technology and
also the relative low uncertainty of
the missions.
From over 1000 miles away with
meter-precision accuracy, the Toma-
hawk missile, a hybrid automated/
autonomous technology, can deliver
a similar payload to that delivered by
a pilot who is susceptible to fatigue
and potentially being shot down.
Because of its digitized scene map-
ping, a form of computer vision that
is highly reliable and does not use
machine learning, the accuracy of
the Tomahawk missile is extremely
high, much better than a pilot. In
the case of targeting static build-
ings and other targets of similar cer-
tainty (like bridges, roads, etc., that
are well mapped), today's computer
targeting technology far exceeds

24

mation, or picks the wrong target.
Thus, any ban on such technology
rejects not only our current use of
such technologies, but also the pos-
sibility that future advances could
result in autonomous offensive
weapons that make fewer mistakes
than humans.
Such advances are not near-term
and will be rife with difficulties. The
next section discusses a proposed
different path that seeks to increase
accountability for both engineers
and military decision makers in the
design and operation of these sys-
tems, as well as to improve the safety
of autonomous systems in general.

the capability of humans, especially
those in a cockpit close to the action.
This precision is what drives the
term "surgical strike" and is why the
United States prefers to use these
weapons instead of humans. Howev-
er, such missions are very labor and
time intensive in the planning stages
and the missiles cost approximately
U.S.$1million per copy.
Some weapons that are launched
against static targets have cameras
that display imagery until the point
of impact, which gives the pilot a
small window of time to potentially
update the aim point, or even redi-
rect a weapon off a target if some
problem emerges such as the real-
ization that the wrong target is
about to be bombed due to wrong
coordinates. This could be another
definition of meaningful human
control. However, all the same prob-
lems exist in that these situations
are extremely time-pressured, so
decision biases, stress, and the fear
of physical threat still can dramati-
cally influence pilot performance.
Unmanned Aerial Vehicle, a.k.a.
drone, missions help alleviate this
problem by removing the human
from direct physical threat. Howev-
er, decision biases and the inability
for humans to reliably assimilate all
relevant information in a short peri-
od of time also affect these opera-
tors. Putting military operators into
a crucible of time pressure, over-
whelming volumes of information,
and life and death decisions in the
fog of war is very far from meaning-
ful human control.
In short, autonomous offensive
weapons already exist now that
can reliably hit static, predesignat-
ed, and well-mapped targets. They
are also fire-and-forget weapons,
which are beyond MHC, at least in
real time. They have replaced error-
prone humans and only make mis-
takes when a human incorrectly
either enters the targeting infor-
IEEE TECHNOLOGY AND SOCIETY MAGAZINE

Focus on Meaningful Human
Certification
Whether discussing autonomous or
human-controlled weapons, MHC is
ill-defined and is possibly a red her-
ring. The fog of war, time pressure,
and the increasing amount of imper-
fect information available to warfight-
ers on a battlefield all combine to
make real-time decision making
extremely difficult with high likeli-
hoods of error. The most meaning-
ful form of human control in the use
of offensive weapons is deciding a
priori which targets will be prose-
cuted and under what conditions.
In terms of offensive autonomous
weapons, there are effectively two
layers of target identification. The
first is the human strategic layer,
where a decision maker, typically at
a high level, determines that a tar-
get (which could be a building, per-
son, etc.) should be prosecuted in
accordance with mission objectives.
Such a decision must abide by prin-
ciples of proportion and distinction
within the Law of Armed Conflict
framework. At this strategic layer,
when a target is passed to lower
level military personnel to prosecute,
humans are certifying that the target
is legitimate and the conditions of
legitimacy. Obviously as in the case
∕

DECEMBER 2019
IEEE Technology and Society Magazine - December 2019

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2019
