8.14 Building on Reputation
165
useful recommendations in exchange, as we know from platforms such as Amazon,
eBay, TripAdvisor and many others. Wojtek Przepiorka has found that such recommendations are beneficial not only for users, who tend to get a better service, but
also for companies.
26 A better reputation allows them to sell products or services at
a higher price. Many hotels, for example, use their average score on TripAdvisor as
a key selling point.
8.15 A Healthy Information Ecosystem by Pluralistic
Social Filtering
How should reputation systems be designed? It is certainly not good enough to leave
it to a company to decide, how we see the world and what recommendations we get.
This promotes manipulation and undermines the “wisdom of the crowds”, resulting
in bad outcomes.
27 It is important, therefore, that recommendation systems do not
reduce social diversity. Moreover, we should be able to look at the world from our
own perspective, based on our own values and quality criteria. Otherwise, we may
end up trapped in what Eli Pariser (*1980) calls the “filter bubble”.
28 In such a
scenario, we may lose our freedom of decision-making and our ability to communicate with others who have different points of view. In fact, some people believe that
this is already the case and one of the reasons why political compromise between
Republicans and Democrats in the US has become so difficult.
29 As a consequence,
conservatives and liberals in the US consume different media, interact with different
people, and increasingly use different concepts and different words to talk about the
same subjects. In a sense, they are living in different, largely separated worlds.
Clearly, today’s reputation systems are not good enough. They would have to
become more pluralistic. For this, users should be able to assess not just the overall
quality, which is typically quantified on a simple five-point scale or even in a thumbsup-or-down system. The reputation systems of the future should include different
facets of quality such as physical, chemical, biological, environmental, economic,
technological and social qualities. These characteristics could be quantified using
metrics such as popularity, level of controversy, durability, sustainability, and social
factors.
It would be even more important that users can choose among diverse information filters, and that they can generate, share and modify them. I call this approach
“pluralistic social filtering”.
30 In fact, we could have different filters to recommend
26 Przepiorka [19].
27 Lorenz et al. [20].
28 Whereby we are fed a meager informational diet based on a small subset of the Internet fitting
our tastes (and perhaps also of those who do the filtering), see Pariser [21].
29 Andris et al. [22].
30 Such a system has been implemented, for example, in the Virtual Journal, see https://web.archive.
org/web/20150910001019/http://vijo.inn.ac/.
165
useful recommendations in exchange, as we know from platforms such as Amazon,
eBay, TripAdvisor and many others. Wojtek Przepiorka has found that such recommendations are beneficial not only for users, who tend to get a better service, but
also for companies.
26 A better reputation allows them to sell products or services at
a higher price. Many hotels, for example, use their average score on TripAdvisor as
a key selling point.
8.15 A Healthy Information Ecosystem by Pluralistic
Social Filtering
How should reputation systems be designed? It is certainly not good enough to leave
it to a company to decide, how we see the world and what recommendations we get.
This promotes manipulation and undermines the “wisdom of the crowds”, resulting
in bad outcomes.
27 It is important, therefore, that recommendation systems do not
reduce social diversity. Moreover, we should be able to look at the world from our
own perspective, based on our own values and quality criteria. Otherwise, we may
end up trapped in what Eli Pariser (*1980) calls the “filter bubble”.
28 In such a
scenario, we may lose our freedom of decision-making and our ability to communicate with others who have different points of view. In fact, some people believe that
this is already the case and one of the reasons why political compromise between
Republicans and Democrats in the US has become so difficult.
29 As a consequence,
conservatives and liberals in the US consume different media, interact with different
people, and increasingly use different concepts and different words to talk about the
same subjects. In a sense, they are living in different, largely separated worlds.
Clearly, today’s reputation systems are not good enough. They would have to
become more pluralistic. For this, users should be able to assess not just the overall
quality, which is typically quantified on a simple five-point scale or even in a thumbsup-or-down system. The reputation systems of the future should include different
facets of quality such as physical, chemical, biological, environmental, economic,
technological and social qualities. These characteristics could be quantified using
metrics such as popularity, level of controversy, durability, sustainability, and social
factors.
It would be even more important that users can choose among diverse information filters, and that they can generate, share and modify them. I call this approach
“pluralistic social filtering”.
30 In fact, we could have different filters to recommend
26 Przepiorka [19].
27 Lorenz et al. [20].
28 Whereby we are fed a meager informational diet based on a small subset of the Internet fitting
our tastes (and perhaps also of those who do the filtering), see Pariser [21].
29 Andris et al. [22].
30 Such a system has been implemented, for example, in the Virtual Journal, see https://web.archive.
org/web/20150910001019/http://vijo.inn.ac/.