Tim Berners-Lee
Date: 2018-01-13, last change: $Date: 2023/03/23 18:37:04 $
Status: personal view only. Editing status: Good enough. Started 2018, but revisited in 2023 as Solid needs the concept. The word Beneficent (very close to Benevolent) came up in MIT's rules about experiments with Human Subjects: one of the criteria for doing work involving people, apart from making sure they were informed and so on, was that the overall work was in they interest: you were trying to cure disease, not generate it, and so on. The concept of a User Agent is not all strange to the web - in fact that is what a browser is called.

Up to Design Issues

Beneficent Apps

It is a sign of the times (2010's) that we even have to talk about these. Back in the days of floppy disk based PCs, when you would spend your money on a cool program (or App as we say now) you would spend money to be able to do useful, fun things. Play a game. Fly a plane. Write an essay. Do your taxes. You would typically have the program in the A: drive and put a disk for your data in the B: drive.

The data on the B: drive was completely in your control. You could use it with different programs in the A: drive. The program in the A: drive was our tool. It helped you do your work. It worked for you.

Until, that is, for me one day when Quicken, the program I had bought ages ago to do my finances and taxes, when it had done my taxes, asked words to the effect of "Are you sure you have enough insurance? Would you like to buy some great insurance?"

That was the end of an era: the era in which I trusted Quicken to be my representative and work on my behalf. Of course millennials are so used to doing everything on the web, using web-based tools, and so used to those web-based tools actually working in someone else's interest, that they may assume this is normal, and take this as the default.

But in fact in this world we loose something very important. The basic human ability to use a computer provides a wonderful level of empowerment. There is something important about a program which represents me.

While Beneficent Apps are not the norm on the web, or even on mobile devices, they in fact common. Every open source app is (should be!) a beneficent app. These are apps which are developed by a community for its own use, and generally they are developed with the needs and wants of the end user in mind.

Web browsers are important Beneficent Apps. They are crucial as the tool with which the person interacts the web and all the crazy wonderful stuff out there In the HTTP protocol spec, the browser is in fact called the User Agent. Web browses must protect and the serve the user in lots of ways


When designing a Beneficent App, always just think at each design choice -- what would the user want the app to do? If you are thinking of yourself as a main driving user, then think about use cases which empower, and connect to other powerful things you can do or will be able to do. But also think of users who are differ from you in many ways - their level of tech ability, their preferences about being social or not, their situation, their personality.

If you have an income stream from selling the data from your users, then you are not likely to build a Beneficent App by default.


How to you measure how good is a beneficent app? It is so easy to make metrics for non-beneficent apps: the engagement level, click-through, the ad revenue or sales revenue they provide. It is more difficult to measure how useful you have been to your user. A user may just be really well informed of something really important, but not do anything which your app could pick up. So yes you can survey them, but now lets's look at measuring their activity. If your app is a something which helps people organize parties, then you can measure the number of parties which people organize. (not to mention whether they were great parties :-).. but those end goals come infrequently, so you could measure the amount of stuff going on to the end of the end goal: the amount of chat (how about sentiment analysis of the chat as to whether it is happy, constructive?) the extent to which the to-do system works - do tasks get done by different people to the people who raise then, for example -- once think of lots of potential things you could measure which may or may not be useful. Then when you have that list, you can look at previous parties and see ho they were involved with making great parties, or with actually getting the party organized at all (which of those would you want to optimize for?) You can also try your apps out, doing A/B testing in the next version as things to optimize for. But also beware of unexpected negative effects. Did they people building social networks imagine the effects on teenage health of a beauty based economy? Probably not, but now we know that systems build to be happy centers of collaboration can end up being toxic for classes of users. To be beneficent, you have to also do no harm!

Rugulating Agents

The concept of apps which work for you, the concept of something which is your agent, is not in fact foreign to our current world. We have it, after all with doctors, and with lawyers. A doctor takes the Hippocratic Oath, or some form of it, which they commit to operate in the interests of the patient. Lawyers also are bound to put the interest of their client first.

There are therefore a lots of laws and regulations out there. to take inspiration from when wording commitments which Apps, or the developers which create them, if anyone wanted to craft regulations ot terms and conditions about Beneficent Apps

Beneficent AI

As AI gets more powerful, every step it takes it becomes more important that it is beneficent. Beneficent for you the individual, and for us the human race. More important that you have AIs which work for you not someone or something else.

Hey online Ad system, Who do you work for? When you recommend I eat at a restaurant, is that the best one for, me or a the result of an instant online auction for who ever bids most to you for my custom? Hey, Siri, who do you work for? Hey, Alexa, who to you work for?

Can you even imagine an AI that works for you? I can, and he's called Charlie