Currently there is a serious trend about new types of collaboration between humans and machines to solve problems.
The information overload: before 100% solved by computer, now 80% computer/ 20% human
Before a lot of hard problems were solved only by computer. Humans were considered only as end-users. Since the democratisation of Internet and the integration of the computer in our life, the interaction between the computer and the human seems easier, more natural. The computer is no longer considered as a black box, but more as a white one with emerging and tight collaboration with people: people can interfer during the computer’s process to help it to solve human problems.
The field of information retrieval is maybe the most obvious case. One decade ago, keyword-based search engines had could crawl all the web and presented few results according to the user’s keywords. Their ranking model based on keywords and content analysis worked quite well due to the information scaricity. But the hyper-growth of the volume of information, mixed with malicious hackers/companies spamming the system flooded and seriously degraded the relevance of such model. In response to this problem we saw the development of global recommendation algorithms using hyperlinks in the webpage as an implicit voting mechanism done by the people, giving a new kind of relevance (reputation) to the page. The most famous one is the pageRank algorithm developped by Google. More recently this trends is getting more visible with the web2.0 phenomena and the new social collaboration/filtering mechanisms in information retrieval. With these new large-scale mechanisms, the user now explicitly votes (e.g. digg.com).
But other fields follow this trend. An exotic example is in evolutionary computation  with the development of a new technic called “human based genetic algorithm”  viewing the algorithm as a business organization, with some outsources done by humans. Futhermore new fields begin to emerge like human based computation for instance.
The social interaction overload: before 100% solved by human (secretary :-), now 80% human / 20% computer
But the reverse should also be true.Even if the user helps the computer to solve the problem of information overload with new mechanisms, the social interaction overload is still a problem.
Due to the communication convergence (i.e. interoperability between the different media of communication) and the social trend of Internet, it’s easier to meet new peoples and interact with them. The time dedicated to our communication has so increased considerabily (How long do you spend to communicate with people by email, IM, phone per day?)[6,7,8].
But, contrarily to the information overload problem, which is originally a problem solved by the computer (e.g. how finding the best information or filtering the information), the interaction overload is originally more a human problem that we try to solve each day (how choosing the best interaction or limiting the user’s solicitations).
The current main communication tools, email, phone, Instant Messenger (IM), have some basic mechanisms to handle this kind of problem: for the phone a voice mail to not be interrupted, idem for the IM: an online presence status (busy, connected, not available). For the email only simple notification or basic rules to classify our emails.
If these mechanisms were enough to solve the interaction overload, why does manager, a type of people who is strongly faced to this problem, still hires a secretary as, inter alia, a mediator to manage this problem? it’s because these mechanisms are too basic and the managers need a higher level of expertise/intelligence to manage it.
As a secretary, the computer should know what are your current tasks/interests/activity, and so with who you want communicate prioritary. It could then choose the most appropriate medium if someone try to interact with you according to our current situation (e.g. if you receive an important email but you’re offline, it sends you an sms, or create a VOIP call with skype between the online user and your cellphone) but also according to the type of interaction needed (Media Synchronicity Theory ). It can also filter the people, or just queue them according to the need you have to interaction with them, etc.: i.e all the tasks of a tradionnal secretary
New tools , using more complex mechanisms than the current case and integrating maybe social/ cognitive/ psychological theories, should emerge creating real artificial mediators between me and the others, and allowing to manage more intelligently the user’s social attention.
 pageRank , wikipedia source
 Human-based_computation , wikipedia source
 Talk at google TechTalk Conference about the works of Luis von Ahn about human based computation (video)
 Evolutionary computation, wikipedia source
 Kosorukoff & Goldberg “Evolutionary computation as a form of organization” (2002)
 A cure for e-mail attention disorder?
 Basex’ report “The Cost of Not Paying Attention: How Interruptions Impact Knowledge Worker Productivity” (pdf)
 “Slow Down, Brave Multitasker, and Don’t Read This in Traffic” The new york times
 “Outcomes from Conduct of Virtual Teams at Two Sites: Support for Media Synchronicity Theory”
 “Interaction Overload: Managing Context and Modality” (pdf)