The Singa­pore Insti­tute of Arbit­rat­ors invited me to debate the fol­low­ing motion: ‘This House Believes That Arti­fi­cial Intel­li­gence Will Have Replaced Arbit­rat­ors with­in Twenty-Five Years’. In short: can – will – algorithms replace arbit­rat­ors with­in a gen­er­a­tion?

We were debat­ing this last night. Here are my open­ing and clos­ing state­ments.

Group picture

There were two debat­ing teams and a tribunal of three judges, Alessa, Jae Hee and Ankit, all peers. Sapna took over mod­er­at­ing at short notice. This meant a good time was guar­an­teed. Katie and Divyesh argued for the motion, Ben­son and I argued against it. In the end, we ‘won’ the debate by the judges’ decision and the audience’s show of hands, but com­pared to a vote at the begin­ning, the pro­ponents of the motion per­suaded many in the audi­ence to change their minds. ‘A mor­al vic­tory on both sides’, as Andy from the SIArb put it (his account of the even­ing is here, by the way).

I guess what I’m try­ing to say is it was great fun.

The Opening Statement

Hallo every­body. Thank you, Divyesh, for your remarks on why you believe arti­fi­cial intel­li­gence will have replaced arbit­rat­ors with­in twenty-five years.

That won’t hap­pen. And why not?

For the same reas­on why a robot vacu­um clean­er isn’t, and won’t be with­in the next twenty-five years, good at foot­ball.

Football

Why do I say that? Don’t a robot vacu­um clean­er and a human foot­ball play­er have a lot in com­mon? For example, don’t they both move in a field which is staked out length­wise and width­wise? As such, isn’t it con­ceiv­able that we devel­op robot vacu­ums which can cov­er the dimen­sions of a foot­ball pitch? Isn’t it con­ceiv­able, then, that we devel­op robot vacu­ums which do so in line with the rules of foot­ball but much faster, more endur­ingly and per­haps even with bet­ter over­view than any human foot­ball play­er?

Yes. But foot­ball isn’t a two-dimen­sion­al game.

In foot­ball, goal­keep­ers dive for saves and field play­ers send over crosses. The ball is thrown in from the side line and kicked through the air where it’s often headed. Some­times, rarely, you see a play­er per­form an over­head kick.

All these actions hap­pen in the third dimen­sion of height. Foot­ball wouldn’t be the same without them. Rather, it’s fair to say these actions con­sti­tute foot­ball. But they are lit­er­ally above any robot vacu­um clean­er that may stray onto the pitch.

Ben­son Mak­ing One of Many Excel­lent Points

Arbitration

AI and human arbit­rat­ors have a lot in com­mon, too. In prin­ciple, both are cap­able of estab­lish­ing both the facts of a dis­pute and the rules applic­able to it, and of apply­ing one to the oth­er. Nowadays the dif­fer­ence between them doing this is merely gradu­al. Thus, isn’t it con­ceiv­able that with­in twenty-five years we devel­op AI which does these things much faster, cheap­er and less error-prone than any human arbit­rat­or?

Maybe. We’re already well on the way. There are self-learn­ing algorithms and all. But it would be wrong to con­fuse a pic­ture detail with the big pic­ture.

The game of arbit­ra­tion isn’t two-dimen­sion­al either. In addi­tion to estab­lish­ing a set of facts and apply­ing a set of rules to them, there’s the hand­ling of the parties and their coun­sel through­out the pro­ceed­ings and cer­tainly dur­ing the hear­ing. There’s the read­ing between the lines, the know­ing or at least sens­ing the exist­ence of cul­tur­al idio­syn­crasies. The cap­ab­il­ity to innov­ate, not just to emu­late. To grasp con­text. Be empath­ic. Have a sense of irony, where applic­able.

All these cap­ab­il­it­ies and skills hap­pen in the dimen­sion of con­scious­ness. Con­scious­ness is the abil­ity to feel things: joy, pain, love, anger. Arbit­ra­tion would not be the same without it. Rather, it’s fair to say con­scious actions make up arbit­ra­tion. Con­scious actions, how­ever, are above any AI.

Three-Dimensional Robot Football

Of course we could set our minds on devel­op­ing robots which play three-dimen­sion­al foot­ball. Even then, not many of us would have the cheek to say robots would replace foot­ball play­ers with­in twenty-five years. It’s just too chal­len­ging to bring the loco­motor sys­tem of robots on par with ours. But we could devel­op these robots and improve them con­tinu­ously if we wanted to, and why? Because in addi­tion to length and width we fathom the third dimen­sion of height.

Three-Dimensional AI Arbitration?

On the oth­er hand, we seem to be incap­able of cre­at­ing arti­fi­cial con­scious­ness. That’s because we don’t get con­scious­ness ourselves. The under­stand­ing of con­scious­ness is not only above AI, it’s also above us, remark­ably.

Patrick Dahm speakingFor dec­ades we have been con­cep­tu­al­ising and for many years we have been devel­op­ing AI. We are improv­ing it now, at dizzy­ing speeds and with fas­cin­at­ing res­ults. But in doing so we haven’t elev­ated AI one iota into the sphere of con­scious­ness.

Nev­er­the­less, AI beats us at Jeop­ardy, chess and Go now. It’s really smart. Which is why arbit­rat­ors ought to use AI and ought to use it more as it devel­ops – don’t get me wrong on this!

AI has become so smart, if we had put the same effort into devel­op­ing foot­ball-play­ing robots, they would be amaz­ingly fast in going after the ball, unri­valled per­haps at game tac­tics and bril­liant in anti­cip­at­ing off­side pos­i­tions. We might have seen the rein­ven­tion of low-pass play. But not many matches would have been decided by such a robot.

It seems reas­on­able to con­tin­ue improv­ing AI as tools, as tools for arbit­rat­ors and oth­er dis­pute resolv­ers as well. After all, AI is good at sort­ing and struc­tur­ing dis­pute-rel­ev­ant data, and fast, and it can help us pre­dict the prob­ab­il­ity of case out­comes.

But if we indulge in the fas­cin­a­tion for AI and replace arbit­rat­ors with algorithms, the res­ult will be a quick­er, here and there even improved, and yet poorer form of arbit­ra­tion. Without sense for nuances or pecu­li­ar­it­ies, without grasp of con­text. Without cap­ab­il­ity to innov­ate, only with the abil­ity to emu­late. Lots of IQ, but no EQ.

Like foot­ball played in two dimen­sions only.

That’s why this won’t hap­pen with­in twenty-five years, neither in foot­ball nor in arbit­ra­tion.

The Closing Statement

Dis­pute res­ol­u­tion by AI seems to be a threshold issue in that it will require of algorithms some­thing which, as of today, they don’t seem to mas­ter.

Nowadays AI can be a tool.

To become a dis­pute resolv­er by itself, AI would have to be cap­able of com­pre­hend­ing both the facts of a dis­pute and the rules applic­able to it com­pletely, and of apply­ing one to the oth­er. This would be the easi­er (I’m not say­ing: easy), the easi­er describ­able the facts and rules applic­able were.

But often they’re not. There are fuzzy things like ‘appro­pri­ate interests’ or ‘con­text’.

Fuzzy Things

An example, by Eng­lish cog­nit­ive sci­ent­ist Mar­garet Boden: a man and a woman go out into the forest to pick blue­ber­ries. He’s twenty years old and can pick ten pounds of ber­ries per hour, she’s eight­een and picks eight pounds an hour. How many pounds will they bring home an hour later?

The answer of course: maybe zero.

Try to get that answer out of an algorithm!

There are vague leg­al con­cepts, too. And there isn’t pre­ced­ent for everything. If there’s a dis­pute to resolve but no pre­ced­ent to be found, what to do?

We Get the One, We Don’t Get the Other

Judges and arbit­rat­ors of today deal with these impon­der­ables by apply­ing (i) their intel­li­gence and (ii) their con­scious­ness.

We’ve made a lot of head­way in arti­fi­cial intel­li­gence. We haven’t made sim­il­ar pro­gress in arti­fi­cial con­scious­ness. Of course, per­haps con­scious­ness as an ele­ment of resolv­ing issues could be sub­sti­tuted with some­thing else. But we are even more clue­less about what that could be.

Reason

But why isn’t it enough to resolve a dis­pute using intel­li­gence? Why the need for con­scious­ness? We sub­mit, because humans like their dis­pute res­ol­u­tion not to be cor­rect and just alone – whatever this may mean –, but to be reasoned also.

Reas­on, how­ever, is the capa­city for con­sciously mak­ing sense of things, estab­lish­ing fact and veri­fy­ing facts, apply­ing logic, and chan­ging or jus­ti­fy­ing prac­tices, insti­tu­tions, and beliefs based on exist­ing or new inform­a­tion. In short, the capa­city to apply to a giv­en situ­ation everything a grown-up human has learned in her life.

Com­pared with this, the task of, say, ana­lys­ing a hand­ful of non-dis­clos­ure agree­ments or keep­ing track of a few dozen Go stones is simple.

Anoth­er example, some­thing I read in the Kluwer Medi­ation Blog the oth­er day: all of us under­stand when someone says ‘a man walks into a bar’. We all know this is the format of a cer­tain type of joke.

Also, all of us under­stand when someone says ‘a man walks into a Saudi Con­su­late’. We are imme­di­ately reminded of the news con­text of only a few weeks ago. We all know this isn’t a joke.

But will an algorithm be aware of this con­text?

We assume parties who choose arbit­ra­tion want their dis­pute resolv­er to be able to tell one from the oth­er.

Different Ball Game

But sure, the day may come when a self-learn­ing algorithm will be intel­li­gent, albeit not con­scious enough to resolve a dis­pute. If we tell it to do so, it will also dis­close the steps of its decision-mak­ing. We might not be able to make sense of it, if it’s a neur­al algorithm which pro­cesses pat­terns asso­ci­at­ively, but that’s a dif­fer­ent issue. In any case, the algorithm won’t be able to reas­on it out.

But maybe, with­in twenty-five years from now or whenev­er, there will be parties, arti­fi­cial or not, who put no value to reasoned decisions? Cer­tainly AI could resolve dis­putes among these parties?

Oh, sure.

I’d hes­it­ate to call this way of dis­pute res­ol­u­tion arbit­ra­tion though. Just because it’s a ball game doesn’t mean it’s foot­ball.