We've tried "look out the window." We still have collisions. Since
human capabilities change only over evolutionary time, and training
programs that encourage good use of existing capabilities have been in
place for some time, we should assume that training and exhortations
have achieved as much improvement as they ever will. The remaining
collision risk must be reduced through some other means.

The traffic pattern and thermals are two high-density traffic
environments where aircraft maneuvering renders collision prediction
difficult. It's not just difficult for machines, it's also difficult
for pilots. During pilot training the task must be taught in several
steps: 1) be aware of how many other aircraft are nearby 2) locate
them 3) avoid getting close to those aircraft unless necessary 4) if
proximity is necessary, watch (i.e., try to predict) the path of the
other aircraft and avoid going toward the place where it is going 5)
learn to anticipate possible unpredictable variations in the path of
the other aircraft also and avoid going toward those areas. Level 5)
is probably only required in thermals and in formation flying. When
we begin thermaling, most of us have to use 3) because we're not good
enough at 4) or 5).

However, at the moment pilots are much better than machines at 4) and
5), while machines are much better than pilots at 1) and 2). Yet, if
1) fails, the rest is useless.

The fact that machines can't do the whole job does NOT mean they can't
be helpful. A machine that could inform a pilot that there are 5
other aircraft in the thermal within +/- 500 ft would be valuable to a
very alert contest pilot who could account for only 4 of them. The
tasks of finding the 5th, and avoiding all 5, might still have to rely
on the Mark I Eyeball.

Of course, pilots in gaggles know that they are in a high
collision-risk situation, and they devote significant attention to
seeing and avoiding other aircraft. Pilots who think they are alone
in the sky devote much less mental capacity to those tasks. Insisting
on "always" maintaining lookout vigilance is ill-advised: many of us
have a pretty high cognitive load a high percentage of the time in
flight, and if we devote too much attention to lookout we may well
lose navigational or meteorological situational awareness, or even
just tire ourselves out mentally, leaving ourselves vulnerable when
attention is important later. This is where a machine could help, by
maintaining a scan and verifying that, indeed, the collision risk is
low. If that changes, the machine can alert the pilot, allowing the
pilot to properly switch mental capacity to "see and avoid." In fact,
this is the primary benefit of flight following during powered VFR
flight, and it's no small benefit.

Insisting that a technology is useless unless it can solve the whole
problem makes perfection the enemy of the good. It's also, in this
case, blind to the imperfection of the current technology - the Mark I
Eyeball - which plenty of science has shown is, in most near-miss
scenarios, far less valuable than the sheer size of the sky.