Recently Stephen Downes linked to an article on the various levels of sophistication of AI personal assistants (by … and …). He added that while all efforts are currently at the third level of those 5 he sees a role in education for such assistance only once level 4 or higher is available (not now the case).

AI assistants maturity levels

Those five levels mentioned in the article are:

  1. Notification bots and canned pre-programmed responses
  2. Simple dialogues and FAQ style responses. All questions and answers pre-written, lots of ‘if then’ statements in the underlying code / decision tree
  3. More flexible dialogue, recognising turns in conversations
  4. Responses are shaped based on retained context and preferences stored about the person in the conversation
  5. An AI assistant can monitor and manage a range of other assistants set to do different tasks or parts of them

I fully appreciate how difficult it is to generate natural sounding/reading conversation on the fly, when a machine interacts with a person. But what stands out to me in the list above and surrounding difficulties is something completely different. What stands out to me is how the issues mentioned are centered on processing natural language as a generic thing to solve ‘first’. A second thing that stands out is while the article refers to an AI based assistant, and the approach is from the perspective of a generic assistant, that is put to use into 1-on-1 situations (and a number of them in parallel), the human expectation at the other end is that of receiving personal assistance. It’s the search for the AI equivalent of a help desk and call center person. There is nothing inherently personal in such assistance, it’s merely 1-on-1 provided assistance. It’s a mode of delivery, not a description of the qualitative nature of the assistance as such.

Flip the perspective to personal

If we start reasoning from the perspective of the person receiving assistance, the picture changes dramatically. I mostly don’t want to interact with AI shop assistants or help desk algorithms of each various service or website. I would want to have my own software driven assistant, that then goes to interact with those websites. I as a customer have no need or wish to automate the employees of the shops / services I use, I want to reduce my own friction in making choices and putting those choices to action. I want a different qualitative nature of the assistance provided, not a 1-on-1 delivery mode.

That’s what a real PA does too, it is someone assisting a single person, a proxy employed by that person. Not employed by whomever the PA interacts with on the assisted person’s behalf.
What is mentioned above only at level 4, retained context and preferences of the person being assisted, then becomes the very starting point. Context and preferences are then the default inputs. A great PA over time knows the person assisted deeply and anticipates friction to take care of.

This allows the lower levels in the list above, 1 and 2, the bots and preprogrammed canned responses and action, to be a lot more useful. Because apart from our personal preferences and the contexts each of us operates in, the things themselves we do based on those preferences and contexts are mostly very much the same. Most people use a handful of the same functions for the same purpose at the same time of day on their smart speakers for instance, which is a tell. We mostly have the same practices and routines, that shift slowly with time. We mostly choose the same thing in comparable circumstances etc.

Building narrow band personal assistants

A lot of the tasks I’d like assistance with can be well described in terms of ‘standard operating procedures’, and can be split up in atomic tasks. Atomic tasks you can string together.
My preferences and contextual deliberations for a practice or task can be captured in a narrow set of parameters that can serve as input for those operating procedures / tasks.
Put those two things together and you have the equivalent of a function that you pass a few parameters. Basically you have code.

Then we’re back to automating specific tasks and setting the right types of alerts.

Things like when I have a train trip scheduled in the morning, I want an automatic check for disturbances on my route when I wake up and at regular intervals until 20 mins before the train leaves (which is when I get ready to leave for the rail way station). I want my laptop to open up a specific workspace set-up if I open my laptop before 7 am, and a different one when I’m re-opening my laptop between 08:30-09:00. I want when planning a plane trip an assistant that asks me for my considerations in my schedule what would be a reasonable time to arrive at the airport for departure, when I need to be back, and I want it to already know my preferences for various event times and time zone differences w.r.t spending a night before or after a commitment at the destination. Searching a hotel with filter rules based on my standard preferences (locations vis-a-vis event location and public transport, quality, price range), or simpler yet rebook a hotel from a list of previous good experiences after checking if price range e.g. hasn’t changed upward too much. Preference for direct flights, specific airlines (and specific airlines in the case of certain clients) etc. Although travel in general isn’t a priority now obviously. When I start a new project I want an assistant to ask a handful of questions, and then arrange the right folder structure, some populated core notes, plan review moments, populate task lists with the first standard tasks. I only need to know the rain radar forecast for my daughter’s school start and finish, and where my preferred transport mode for an appointment is bicycle. For half a dozen most used voice commands I might consider Mycroft on a local system, foregoing the silos. Keeping track of daily habits, asking me daily reflection questions. Etc.

While all this sounds difficult when you would want to create this as generic functionality, it is in fact much more simpler in the case of building it for one specific individual. And it won’t need mature AI natural conversation, merely a pleasantly toned interaction surface that triggers otherwise hard coded automated tasks and scripts. The range of tasks might be diverse but the range of responses and preferences to take into account are narrow, as it only needs to pertain to me. It’s a narrow band digital assistant, it’s the small tech version.

Aazai

For some years I’ve dubbed bringing together the set of individual automation tasks I use into one interaction flow as a personal digital assistant ‘Aazai’ (a combination of my initials A.A.Z. with AI, where the AI isn’t AI of course but merely has the same intention as what is being attempted with AI generally). While it currently doesn’t exist mostly as a single thing, it is the slow emergence of something that reduces digital friction throughout my day, and shifts functionality with my shifts in habits and goals. It is a stringed together set of automated things arranged in the shape of my currently preferred processes, which allows me to reduce the time spent consciously adhering to a process. Something that is personal and local first, and the atomic parts of which can be shared or are themselves re-used from existing open source material.

Last week I have made changes in the way I process email. Adapting it more towards ‘Getting Things Done’, which I had avoided doing for years, and making some changes in daily habits around it. Now that I made the change, I can’t quite understand what kept me from doing it, even though before I thought my needs would not be met with a new routine.

Resisting change
As I use Gmail and have the notion I really shouldn’t, I mentally postponed changing my routines until ‘after replacing Gmail’, assuming I would otherwise either increase the cost of leaving Gmail (by having routines more deeply connected to its functionalities), or I’d find a replacement that already contained a better flow by default, making designing change now a waste of time. I know, neither make much sense upon closer consideration. Likely the real reason I made the change now, is having come back from a long vacation, and not many obligations yet as most clients are still away themselves. That, and receiving an external trigger right at that moment.

The trigger
That trigger was getting a message from Martin Roell, that one of his colleagues, Rob van den Brand, is offering a free download of how to deal with email. I downloaded it ouf of curiosity to see if it contained anything new in terms of suggestions. At first glance it didn’t, it was the GTD style approach I already knew (using the 2 minute rule, sorting into piles to reply, read, do and other etc.). Then a few days later a follow-up arrived with a few behavioral tactics to help make the mechanism work. The first one was unsubscribe to a lot of stuff, which led me to review my automated filtering, which led to re-evaluating the original GTD method, which led to implementing it……

The old routine
Over the years I kept all my e-mail in my inbox, always. Piling, never filing (tagging I do). The original reason for that was that my first mobile e-mail app would not let me easily access and search archived mail, only what was in my inbox would be readily available. The first step of my mail process is usually on my mobile.

From the newly arrived mail, I would ‘star’ those I thought would need some sort of follow-up. Things that don’t interest me I would leave unread but not throw out. This would be my basic triage method. I also have various filtering rules that label incoming mail (apart from ‘starred’) according to what part of my professional activities they represent (my company, my fablab stuff etc.) Using Gmail’s multiple inbox feature, those stars and labels were presented on my laptop screen as separate lists next to the main inbox.

At some point during the day I would open my mail (Gmail’s web interface) on my laptop and:

  • mark all remaining unread mail as read
  • work through the starred items
  • look at/answer mail while working on a specific professional context (1 of the inboxes)

The problem with this was that a starred mail could still mean many things (migh be interesting, immediate action, little or lots of work, read, keep in mind etc.) I basically needed to reevaluate every single mail, every time I opened up the ‘starred’ list. Over time that list would grow with unprocessed items from the past, becoming a ineffective mental drag, except for the recently starred messages. Also some of the multiple inboxes had survived beyond their waned usefulness due to changing focus and activities, and I had difficulty putting them to new good use.

The new routine
My current mobile app (the place where I do my first mail triage) fully supports labeling messages and accessing archived mails. Functionality I wasn’t putting to good use. So that makes it possible to do more detailed and better triage on incoming mail. I now, following the GTD material I mentioned above:

  • use many more filtering rules to automatically process and label incoming mails, alerts, mailers etc.
  • have unsubscribed a wide range of mailers connected to long time ago interests
  • have moved some quarter million mail exchanges of the past years from the inbox to the archive
  • label the remaining few mails that still reach my inbox with 1_reply, 2_todo, 3_toread, 4_waiting and other assorted relevant labels (such as ‘bookkeeping’, ‘opendata’, ‘acquisition’ etc.) so they can be more easily found back when needed
  • create new filtering rules when a mail arrives that warrants a filter
  • empty the inbox by moving all labeled and remaining unlabeled mail into the archive

The original multiple inboxes I now show below new mail, in stead of to the right where I had them for the past years. The multiple inboxes now show the reply/todo/read/waiting labels. That looks like this:

mailinbox

Key take-aways and needs
Changing my mail process and method of triage turned out to be easy. It moved the decision what to do with an e-mail forward 1 step, and made it part of the triage. (Before I would only star a mail and then decide later). This makes my normal daily time slot for mail sufficient to actually deal with the contents of that mail.

Main ‘win’ is that my mail interface is much less noisy, both due to heavier automated filtering and removing processed messages from the inbox. Before I would see whatever was left over from ‘before’ and always have a full page of messages in front of me. That clutter is now 5 short lists, with only one of those lists needing attention at any given time. All other stuff from mailing lists are available under a label/tag, when I decide I want to catch up but never clog up the inbox.

My main demand, being able to do triage ‘on the go’, is still being met (and more automated than before). The reduced clutter also feels like it might be a benefit when I move out of Gmail.

The only thing to still do is to much better connect the list of mails labeled to-do to my actual task management tool (Things, by Cultured Code) and making sure they get the right follow-up that way. I could probably automate that, but haven’t figured out how to do that yet. This may mean that the to-do part of the mail flow will actually disappear from my gmail altogether.

Working with a Getting Things Done system in the past 9 months or so leads to a few thoughts I’d like to share.
Because it seems to me there is a systemic weakness in the concept of GTD. This does not mean GTD is not bringing me benefits, on the contrary. It does limit its scope of effectiveness though.
GTD, what it does
GTD is about making lists, more effective lists, to manage your time/life better.
The biggest benefit in GTD, as I understand it, is in not asking you to attach priorities or times to activities in your list, as time management systems generally do. It assumes that once you have good lists you will know what to do, based on time and energy available, as well as your own sense of urgency. This is a true diamond, as it trusts you to be human, and doesn’t demand the conveyor belt mindless behaviour other time management systems ask for (“once you have the right list, you’re on mindless autopilot”)
The other big benefit of GTD is its multiple feedback loops. The short one, shared with other time management systems, informing you about tasks, and tasks that are waiting for someone else. The longer feedback loop(s), the reviews, allow you to step away from the task units, and look at your goals if they are still valid, and if your tasks still serve those goals. This helps you prevent to be running because you are busy, without knowing why you’re busy and what it’s all for. Doing good reviews (both back and forward looking, so review is a partly misleading term), and doing them regularly however is not easy.
GTD, what it does not
The biggest problem of GTD is that it is based on lists. Because list making is an old and time-honoured information strategy. GTD in essence says: if your inbox and the amount of tasks is growing and your life is getting more complicated make better lists.
That amounts to, when someone does not understand you, repeating yourself saying it LOUDER. In stead of choosing different words to convey your message. GTD is trying to apply the list making strategy better, in response to a failing list making strategy.
However when I see what I and others are trying to do with GTD it is navigating an increasingly fragmented and complex environment. The root causes are quantitative rises in the connections between people (small world), the speed of change (world becoming a metropole), and the amount of information (information abundance). The internet, and other preceding media, as infrastructure play a very big role in these quantitative shifts.
Quantitative changes, qualitative answers
These quantitative shifts are by necessity begetting qualitative answers, because conventional methods (like making lists) stop scaling. Web2.0 tools have some of those qualitative answers (active sharing and sense making, social relations as information filter, networks of meaning) as design principles. Other qualitative answers are becoming part of our information skills (pattern recognition, knowing when to stay focussed amidst distraction, knowledge as being connected/networked, learning as building networks).
I find I apply those qualitatively different information strategies before I can get to the level of things where GTD lists make sense. I hunt for patterns in my RSS feeds, and then those patterns become inbox items. The RSS feed items themselves are not suited to treat as inbox items, simply because the items themselves are not the relevent units of information for me.
I already have marked 90% of my incoming e-mail as read without reading them, before I get to seeing them as true inbox items that warrant a decision to respond to, put on my task list, send to someone else, or delete.
I also find that a very important piece of my work does not get affected by GTD at all: staying aware of my social network and context. Keeping track of the people I know and the communities I am part of is my premier source of learning, of landing projects, of bringing my goals closer, and it is all to a very large extent based on peripheral sense. It is based on not looking directly at it, nor on focussing on it, but glancing at it,. Like the way you keep track of what is happening in a pub by glancing around, while you are actually focussing on the conversation with the person in front of you. Or like the way in the dark you see more out of the corner of your eye, than right in front of you. Like with my RSS feeds this is pattern hunting. And only the patterns I find ever reach my inbox where I focus on them to decide what to do next. Tuning my antennas on my surroundings, and pro-actively define what type of patterns I am currently especially interested in also takes a large chunk of time and energy.
This creates a scope where GTD is effective but only after the problems caused by the size, fragmentation and speed of the world around me have already been dealt with using other strategies. GTD gives me very effective lists, but only after I have created a qualitatively better ‘inbox’ myself. GTD can deal with complicated stuff very well, but I have to deal with complexity myself first.
How GTD could be better
One way in which the GTD method could become more valuable is if I could get patterns from it about what I do, that became inbox items again. Another if I could shape my GTD reviews to help me tune my antennas for the peripheral vision better as I described above. Something to think about further