We ended up working in a kind of calendar/to-do/agenda/personal assistant application that prompts the user to carry out a task in the optimal context in time and space. This application would help the user manage his/her activities. For tasks that have a fixed scheduling, its functionality is limited to providing reminders, exactly as a regular calendar application would. However, for tasks that may be executed freely at any time before a given deadline, it suggests the best moment to do something using information about the nature of the task and the user’s preferences, live monitoring of his/her level of busyness, time, location, etc.

Before I explain how it would work, and because I can picture you shaking your head and tightening your lips, let me say something. This idea sounds familiar to you? Yes, to me too… When talking about applications that are supposed to be smart enough to suggest what you may or may not want to do, there’s a very thin line separating awesomness from complete failure. That’s what, I think, when you present an idea like this to people, makes people stand in two very distinct camps: either you love it or you hate it. There are few things, in UX, as polarizing as agents. I was very sceptical about this idea at first. Then something happened.

I realized I was being guilty of the sin of ninetiness. Was it possible that my Clippy PTSD was blurring my vision? So I tried to shake off the clippiness, and think about it in an objective way. Surely I had seen other (working) examples, but where? I know what I’m going to say may be controversial and that some people will think that I’m abusing the definition of agent. But what does actually happen every time you do a search on Google, for example? That’s no plain text search, and neither is it a text search where results are ordered according to the old version of Page Rank anymore. It’s been a long time now that Google has been taking global and local search trends, current events, your personal profile, previous search queries, search history and it appears that even websites you visit into account when choosing which results to display for your query, they even claim to be able to predict epidemic outbreaks.  According to Wikipedia (hey, I’m not an expert) there are only four essential types of intelligent software agents:

• Buyer agents or shopping bots
• User or personal agents
• Monitoring-and-surveillance agents
• Data Mining agents

If you take a look at the definition for each of them, there’s no category Google search doesn’t belong to. So it looks like this is the secret of successful agents: not looking like a paper clip, actually not looking like anything, looking like they’re not even there. If people like to have their search results tweaked if they still feel that it’s their keyword choice that brought them about, would people accept to have reminders for tasks that they never actually schedule if the system can faithfully interpret their heart’s desires? Can it be done? Well, Google does a pretty good job… so I do think it’s possible. Will it be done? I don’t know, not by me, lol. We just had to come up with a new concept that’d make our(?) lives better, and as much as I came to undertand that this is not such a bad idea, I’m not in love with the concept either, let’s just say I find it interesting.

So, how would it work? The user enters his/her schedule with fixed tasks in the same way as in a normal calendar. Also he/she enters the flexible tasks and some information about them (i.e. deadline, location, duration, sequential constraints, is this task repeating over time). The system uses location information (GPS) to learn about the user’s location, time zone and weather. Locations can be tagged (e.g. office, home, mall). Also, permanent ambient noise monitoring provides information about the current activities and level of busyness of the user (e.g. is he/she maintaining a conversation, typing on the keyboard, walking).
Finally, specific rules entered by the user and machine learning on acceptance or rejection of its suggestions provide the system with the user’s preferences. Users can also manually set states related to mood or desired level of busyness for a certain time period. All these elements are used by the system to suggest the optimal moment to carry out a task based on proximity, weather, mood, availability, etc. This is applicable to management of a daily routine, a weekly schedule and long term planning. Information related to the pending tasks (content feed) and health advice can also help the user make optimal choices about his/her workflow. Finally, casual networking and social commitments are also supported because proximity, availability and coincidence in pending tasks can be monitored.

Advantages, Challenges and Opportunities: this system has the potential to solve the problem of having to remember to look at a deadline calendar or to-do list because the user will be reminded to complete the tasks. Also, the fact that tasks are suggested in the right context will aid efficiency and avoid procrastination. The cognitive load to determine when to perform the task is switched from the user to the application. However this presents some challenges. Users will still have to manually enter information about their tasks, so the suggestions should be useful enough to justify the trade-off. Also, the application should remain non-invasive. This system also allows sharing schedules for joint tasks, displaying the user’s availability status, and the information about the user’s mood, busyness and daily tasks could potentially be used for other applications or to suggest entertainment for the user’s free time, eventually. The wide array of technical capabilities in a mobile phone provide flexible input and output modes that allow to ease task input and minimize invasiveness, respectively. The possibility to implement a multimodal interface also allows to use the system in different scenarios under different conditions.