Script for ICCRTS 2011 Presentation

16th ICCRTS – 22 June 2011

Bruce Forrester a DS from the I2 section in Valcartier. Out teams is working on the
problems of information and cognitive overload in intelligence analysts. Two of the
projects that I am working on, that form the basis of my brief today, are the SNAC – looking
at creating a SNA capability in a COIN context and ASIM in the ASIC context.

Slide – 1 Intro
Trust in relationships is essential for deep sharing of information/intelligence and
meaningful collaboration.

Building the required level of trust is often a lengthy face-to-face process.

Lack of trust seriously hampers effectiveness.

Can leads to redundant analysis and information overload.

Trust-based networks are a very promising avenue.

Trust can be mapped to the digital world, at least partially, through rating,
recommendation and reputation (R3) systems. These systems have proven effective in
online purchasing but can they be adapted to more high-stake endeavours?

Slide – 2 Agenda
• Scenario
• Trust
• Recommendation Systems
• Social Networks
• Challenges
• Conclusion

Slide – 3 scenario

, just two months after the fast-
reaction teams were sent in.

Slide –

As an Intelligence officer, he knew the importance of building good relationships.
Relationships built trust and trust leads to a good flow of information and intelligence.
Apart from leaving his family for six months, he was actually looking forward to this
deployment. During his pre-deployment training, he had participated in a new initiative
that was focused on building trust between intelligence analysts, and the various military
and the non-government organizations that had flocked to the disaster zone to help the
Haitian people.

Slide –

This new initiative was quite a different approach to this problem than his last deployment,
in the southern Afghan theatre, when such a program was not yet in place. He distinctly
remembered the time that fellow soldiers were killed in an operation to rescue a reporter
who had gotten himself into trouble; despite having been warned not to travel into that
particular area, the reporter ignored the advice and was taken by Taliban[9]. After that, it
was hard to convince his troops to show patience with the NGOs.

Slide –

However, he believed that this time, the military could concentrate on achieving their
missions without having to worry about the safety of NGOs. Indeed, the NGOs would be
aiding his own task as they report on activities and participate in a network that enables
information sharing and relationship building.

Slide –

two months ago. Shortly after he received his travel orders he was advised to log into the
METIS network (named after the Titan Goddess of good counsel, advice, planning and
wisdom).

Slide –

METIS was set up to allow individuals from government departments and from non-
government organizations, as well as contractors from industry, to interact prior to and
during deployment to countries in need of aid. The idea was to build relationships through
online social networking that would then translate into trust, or at least better
understanding of one another, on the ground in the theatre of operations. In fact, Major
Jones remembered that he met some of his most important contacts in the communal
coffee garden area in Afghanistan.

Slide –

Over the next two months, he read through the homepage of each of the NGO’s that
provided its missions and objectives and profiled its people. He read the
, as well as many
more who were scheduled to go. He found that some had very similar interests and he was
able to trade some tips on finer points of home brewing. He was able to ask questions and
determine some additional kit that he would need.

Slide –

e. “But as
the past few months have made clear, there is little coordination among the NGOs or
between the NGOs and Haitian officials. Some NGO plans don't fit or clash outright with the
plans of the government. Some are geared toward short-term relief—a classic case of

giving a man a fish instead of teaching him to fish”[10]. Jones was hoping that METIS
would aid in changing this problem.

Slide –

. Luckily, the METIS had a trust-based
recommendation system that allowed him to quickly hone in on the most pertinent
documents as well as the experts in various areas.

Slide –

Now in theatre, Maj Jones was using the METIS system daily to get updates on NGO
movements. As well, he was able to see what other analysts were reading and
recommending. This included all sorts of OSINT (open source intelligence) and HUMINT
(human intelligence) sources as well as NGO situation reports.

Slide – Trust

So let’s go into a few details and considerations for building such a system.

Trust will play a large role if such a social network is to function properly.

There are many characteristics of trust that also need to be taken into consideration when
designing trust algorithms:

a. Trust is dynamic. The degree of trust “b” has in “c” can change over time
depending on the interactions and outcomes between the two.
b. Trust is asymmetrical. The degree of trust “b” has for “c” is not necessarily the
same that “c” has for “b”.
c. Trust has a slow build rate but a quick fall rate.
d. Trust is subjective and personal. “a” and “b” will have different degrees of trust
towards “c” and objective measures are very hard to produce.

e. Composability. There are different paths that could be followed to connect “a”
and “d” (through b & c or through e, f & c).
f. Trust is context-dependent. “a” might trust “b” to provide information about one
country but not about another country.

Slide – Recommendation Systems

The use of recommendation systems aids users in rapidly decreasing the size of the pool
from which to find objects of interest. It essentially acts as a social filter. The algorithms
that sort through user recommendations are usually of two types:

a. those which are based on similarities between the current item of interest and
the items related to it (i.e. a site might show you all books that are related to a
particular breed of dog), or

b. Those based on the similarities of the users’ likes within the system (i.e.
Amazon’s famous – “users that bought this book also bought these books...”), also
known as collaborative filtering.

There are problems with both of these types of filtering.

The first (based on similarities) tends to be impersonal as it does not take into
consideration the characteristic of the user other than what items they have looked at in
the past. In addition, such a system would not be good for finding outliers or emergent
items.

The second, collaborative filtering requires a database of ratings on the items. This implies
that newer items, that are not likely to have many initial ratings, will not be taken in
account by the algorithms. In the intelligent analysts’ case, the most recent information is
critical to the situational awareness and hence collaborative filtering could not be used

without significant improvements. The diversity of information will also need to be taken
into account.

These filtering algorithms tend to recommend only items that are similar. There will need
to be a way to get items that are considered outliers or very different for comparison and
for hypothesis validation.

Slide – Social Networks

Research has shown that “a distributed knowledge system serves to reduce individual
cognitive overload, enlarge the collective pool of expertise, and minimize redundancy” . A
large number of web-based tools could be used to provide a platform for such a pool of
expertise. This platform could take the form of a social network

Realistically, there is far too much data, information, and knowledge out in the world for
any single person to make sense of it, even in a highly specialized area such as warfare.

The “work of the masses – the wisdom of crowds” will be the only way that we can hope to
make sense of it all. .

Slide – Challenges

Some of the questions that I am struggling with:

a. How does one create an online environment that allows for the right mix of these
components of trust such that deep sharing of information can occur?
b. Can the use of personal agents help to create automated trust recommendations?
c. How does the reputation of the organization that one represents affect the level of
individual trust?
d. How sophisticated do the algorithms need to be in order to produce good results?

e. There are many issues to resolve from a human factors perspective. Would
intelligent analysts and NGOs use such a network?

Slide – Conclusion

There is clear evidence that trust-based recommendation algorithms enable users to sort
through vast quantities of information to produce good results, thus decreasing the
information overload of individual users.

However, the current research has concentrated on low-risk subjects such as movies or
opinions on consumer goods.

In the intelligence domain, information takes many different forms consisting of anything
from large academic papers to short situation reports provided by actors on the ground in
an operational theatre. There might be very few recommendations attached to these
artefacts thus limiting the usefulness of collective filtering. Although, perhaps one
recommendation from a highly trusted neighbour would be enough to warrant attention.

While such algorithms might work for some situations, to be useful for intelligence
purposes they would also require a content filter. One might foresee the application of
pattern-matching technology that forms a conceptual and contextual understanding of all
content, independent of language or format.

Combined these two forms of filtering would produce a strong starting point for intelligent
analysts.

Script for ICCRTS 2011 Presentation

More Related Content

What's hot (19)

Viewers also liked (16)

Similar to Script for ICCRTS 2011 Presentation (20)

Recently uploaded (20)

Script for ICCRTS 2011 Presentation