[LINK] Scott Morrison said hate content on social media could be automatically screened out by algorithms. Is he correct? - ABC News (Australian Broadcasting Corporation)

[Roger Clarke]

> On 18/4/19 10:34 am, Antony Broughton Barry wrote: 
>> https://www.abc.net.au/news/2019-04-18/fact-check-can-algorithms-screen-out-hate-content-social-media/10979770 

On 19/4/19 9:30 am, Tom Worthington wrote:
> An algorithm which automatically screened out hate content on social 
> media would likely also filter out many statements by Australian 
> mainstream politicians.

It's an ill wind, etc.  Because maybe we'd then be spared not only a lot 
of pollie-rave, but also a slice of the Question Time farce.


> We need any such mechanism to be imperfect. 

We're pretty safe there.

Has anyone seen any attempts to operationally define 'hate speech'?

Brandis made a complete hash of attempts to reconsider the awkward s.18C 
of the Racial Discrimination Act (Cth).

The key criteria for unlawful speech are to "offend, insult, humiliate 
or intimidate'.

Brandis proposed:
http://www.rogerclarke.com/DV/PFS-1408.html#CC
-   the removal of 'offend', 'insult' and 'humiliate'
-   a definition of 'intimidation' as 'a reasonable likelihood of
     causing of fear of physical harm'
-   the addition of 'vilification' defined as 'reasonable likelihood
     of inciting hatred'

In almost the only instance in which I ever agreed with Brandis about 
*anything*, all of his proposals appeared to me to be well worth 
considering.  (Once I'd analysed the package as a whole, however, I felt 
that the excessively permissive interpretation and saving provision 
undermined the intimidation and vilification protections).

Another existing criterion, along the lines of 'incitement of violence', 
is tricky, but at least you could draft and test some rules for scoring 
passages against that criterion (e.g. multiple occurrences of active 
words associated with violence, such as 'kill' and 'attack';  and 
multiple occurrences of pejoratives relating to a category of people).

But I really can't see things like 'a reasonable likelihood of causing 
fear of physical harm' and 'a reasonable likelihood of inciting hatred' 
being amenable to any vaguely sensible automated application.


> ...  With an effective social 
> media filtering mechanism it will be very tempting for a future 
> government to use it to suppress free speech in Australia.

Absolutely.  Every proposal for censorship has to be considered very 
carefully, and kneejerk reactions vigorously fought.


-- 
Roger Clarke                            mailto:Roger.Clarke at xamax.com.au
T: +61 2 6288 6916   http://www.xamax.com.au  http://www.rogerclarke.com

Xamax Consultancy Pty Ltd      78 Sidaway St, Chapman ACT 2611 AUSTRALIA 

Visiting Professor in the Faculty of Law            University of N.S.W.
Visiting Professor in Computer Science    Australian National University