Should verbs be modified? Speech Acts and Activity Streams...

[Bob Wyman]

As ActivityStreams builds an ever richer collection of supported verbs and objects, it is becoming more and more an effective language for encoding and communicating a broad variety of the locutionary acts or "utterances" that comprise our daily activity and communications. It is unsurprising that much of the focus of this working group has been on the syntax of Activities and the mechanisms for accessing or exchanging them. Yet, the Speech Acts which are encoded as Activities have meaning or "consequences" in addition to syntax. Thus, I think we need to spend a bit more effort considering the semantics of Activities what is meant by the use of verbs. As John Searle would say: "by saying something, we do something."

I am often puzzled when looking at the Activity Stream verbs since I'm never quite sure what exactly is meant by them. I am concerned that while I might use a verb in one sense, it might be understood as having some other sense by one who reads an Activity I created. For instance, consider James Snell's recent discussion of "Flag as inappropriate" using the "report" verb. As others have also noted, there is some ambiguity here that makes it hard for a writer to know what verb to use and makes it hard for readers to know what is the intent of the speaker. The ambiguity comes, I think, because verbs can be used in a variety ways, each having distinct semantics or rhetorical intent, even though there may be little or no distinction between the syntactical aspects of their use. This is a large part of what makes machine language understanding hard. Two syntactically identical utterances can have very different meanings depending on context, etc..

For instance: If I "flag" an object as "innappropriate, which or how many of the following am I doing?

• Asserting that the object is inappropriate (i.e. Simply making a claim?)
• Directing those who receive the Activity to treat the object as inappropriate? (Perhaps I am a manager or someone else who has the authority to tell folk what to do?)
• Committing or promising that in the future I will treat the object as inappropriate? (perhaps I am a site administrator and I'm announcing my policy.)
• Expressing my opinion or feeling that the object is inappropriate. (i.e. I'm asserting something about me and my feelings, not necessarily the object.)
• Declaring that the object is inappropriate, perhaps as the result of some formal review process in which the act of making the declaration has some formal meaning. (i.e. The declaration is performative. It isn't "true" or "false" but rather changes reality as a result of having happened. Another example: "I now declare you man and wife...")

Activity Streams makes a variety of locutionary acts easy to encode. The problem is that the Speech Acts that are represented as Activities cannot be understood based only on their syntactical details. These Speech Acts are also illocutionary acts which have meaning, intent and consequences. What Activity Streams doesn't make clear, as far as I can see, is how a reader and a writer will come to agree not only on the encoding or representation of some verb, but also on what is intended by one who used that verb in a specific Activity.

Similar issues arise with other verbs: 

• Does "like" mean that I am expressing my opinion or is it an assertion that a "Like" button was clicked?
• In an educational context, I may "Assert" that I passed a course, but the authority who provided the course might "Declare" that I have passed.
• If I "disagree" or "reject" is that an expression an assertion or a declaration?

What I'd like to see is a means to explicitly tag or modify the verbs that I use in Activities with their rhetorical, illocutionary intent. Doing so would, I think, greatly expand the expressive capability of Activity Streams at a relatively low complexity cost. I also think it would make sense if, in the documentation of verbs, we could find a "default" classification of each verb.

Does this make sense?

bob wyman

[James M Snell]

On Tue, May 22, 2012 at 8:31 AM, Bob Wyman <b...@wyman.us> wrote:
> [snip]

>
> I am often puzzled when looking at the Activity Stream verbs since I'm never
> quite sure what exactly is meant by them. I am concerned that while I might
> use a verb in one sense, it might be understood as having some other sense
> by one who reads an Activity I created. For instance, consider James Snell's
> recent discussion of "Flag as inappropriate" using the "report" verb. As
> others have also noted, there is some ambiguity here that makes it hard for
> a writer to know what verb to use and makes it hard for readers to know what
> is the intent of the speaker. The ambiguity comes, I think, because verbs
> can be used in a variety ways, each having distinct semantics or rhetorical
> intent, even though there may be little or no distinction between the
> syntactical aspects of their use. This is a large part of what makes machine
> language understanding hard. Two syntactically identical utterances can have
> very different meanings depending on context, etc..

One of the key strengths of Activity Streams lies in the inherent
flexibility and limited ambiguity that exists in the current model.
When I make a statement, "Joe flagged an article as inappropriate" the
general *meaning* of the isolated statement is clear while the
implications of the statement are not. The consequences that occur as
a result of the action are entirely dependent on the context.
Personally, I do not see that as a bad thing. Just as when two people
are speaking, context is generally far more important to establishing
the intent of what is being said than the actual words. By changing
the context, you change the meaning. What's critical, I think, is
finding a way of capturing the original context ... that is, a way of
expressing the Why of a particular Activity in addition to the Who,
What, When and Where aspects we already capture.

>
> For instance: If I "flag" an object as "innappropriate, which or how many of
> the following am I doing?
>
> Asserting that the object is inappropriate (i.e. Simply making a claim?)
> Directing those who receive the Activity to treat the object as
> inappropriate? (Perhaps I am a manager or someone else who has the authority
> to tell folk what to do?)
> Committing or promising that in the future I will treat the object as
> inappropriate? (perhaps I am a site administrator and I'm announcing my
> policy.)
> Expressing my opinion or feeling that the object is inappropriate. (i.e. I'm
> asserting something about me and my feelings, not necessarily the object.)
> Declaring that the object is inappropriate, perhaps as the result of some
> formal review process in which the act of making the declaration has some
> formal meaning. (i.e. The declaration is performative. It isn't "true" or
> "false" but rather changes reality as a result of having happened. Another
> example: "I now declare you man and wife...")
>

It could, of course, be any or all of these depending on the context.
The way Activity Streams is currently defined, implementors are
largely free to make use of the syntax in any way that makes the most
sense to their own specific needs. Those needs will vary, and thus the
use of the syntax will vary. Eventually, over time, we may begin to
notice clear patterns emerge out of the mix of ways the format is
used. Once we begin to spot those patterns, and once we're certain
they truly exist, we can begin to document and formalize those
patterns. Paving the cow-paths as Evan put it.

> Activity Streams makes a variety of locutionary acts easy to encode. The
> problem is that the Speech Acts that are represented as Activities cannot be
> understood based only on their syntactical details. These Speech Acts are
> also illocutionary acts which have meaning, intent and consequences. What
> Activity Streams doesn't make clear, as far as I can see, is how a reader
> and a writer will come to agree not only on the encoding or representation
> of some verb, but also on what is intended by one who used that verb in a
> specific Activity.
>
> Similar issues arise with other verbs:
>
> Does "like" mean that I am expressing my opinion or is it an assertion that
> a "Like" button was clicked?

Why does the specific meaning matter in this case? If "Joe likes
waffles", does it really matter if that means Joe just expressed his
opinion or if he clicked a "Like" button?

> In an educational context, I may "Assert" that I passed a course, but the
> authority who provided the course might "Declare" that I have passed.
> If I "disagree" or "reject" is that an expression an assertion or a
> declaration?
>
> What I'd like to see is a means to explicitly tag or modify the verbs that I
> use in Activities with their rhetorical, illocutionary intent. Doing so
> would, I think, greatly expand the expressive capability of Activity Streams
> at a relatively low complexity cost. I also think it would make sense if, in
> the documentation of verbs, we could find a "default" classification of each
> verb.
>
> Does this make sense?
>

Again, it all depends on the context. I do not necessarily think we
need to modify the verb so much as provide a means of recording the
context of a particular Activity... essentially, why the Activity
occurred and, potentially, the result of the Activity.

For instance:

{
"actor": {...},
"verb": "...",
"object": {...},
"context": {...},
"result": {...}
}

The "context" property would contain information about the context in
which the Activity occurred, providing details about why it happened.
The "result" property would contain information about what happened as
a result of the activity, if anything.

Beyond that, I'm not sure what else may be needed. Do you have an
example of what kind of modification to the verb you had in mind?

- James

> bob wyman
>
> --
> You received this message because you are subscribed to the Google Groups
> "Activity Streams" group.
> To post to this group, send email to activity...@googlegroups.com.
> To unsubscribe from this group, send email to
> activity-strea...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/activity-streams?hl=en.

[Jack Repenning]

On May 22, 2012, at 8:31 AM, Bob Wyman wrote:

> What I'd like to see is a means to explicitly tag or modify the verbs that I use in Activities with their rhetorical, illocutionary intent. Doing so would, I think, greatly expand the expressive capability of Activity Streams at a relatively low complexity cost. I also think it would make sense if, in the documentation of verbs, we could find a "default" classification of each verb.

How do we avoid merely multiplying the collection of ambiguous terms and constructs?

On May 22, 2012, at 9:26 AM, James M Snell wrote:

> Again, it all depends on the context. I do not necessarily think we
> need to modify the verb so much as provide a means of recording the
> context of a particular Activity... essentially, why the Activity
> occurred and, potentially, the result of the Activity.

Again, aren't we merely multiplying the ambiguities?

There seems to be an implicit imitation, here, of the existing model of natural language. But natural language is notoriously ambiguous! As we've all now learned in our "netiquette" training, natural language disambiguation depends heavily on non-verbal side channels: you speak, I look puzzled, you notice and rephrase.

Lacking rich side channels, what we have traditionally done in computer languages is document the definitions. It's laborious, I grant you, and not very trendy (not at all RESTful!), but is there really an alternative?


Jack Repenning

Time flies like an arrow. Fruit flies like a banana.