van Lambalgen, Michiel; Fritz Hamm;
The Proper Treatment of Events
Blackwell Pub., 2005, 251 pages
ISBN 1405112131, 9781405112130
topics: | logic | temporal | language
- DURATION - largely lexical, and not grammaticalized except in rare lgs
that make tense distinctions between "less than one day ago", "one day
ago", and "long time ago".
- TEMPORAL PERSPECTIVE - attitudes w.r.t. past, present future
William James:
[The practically cognized present] [i.e. the specious present] is no
knife edge, but a saddle back with a certain breadth of its own on
which we sit perched.... with a bow and a stem. We do not first feel
one end and then feel the other after it, and from the succession
infer an interval of time between, but we seem to feel the interval
of time as a whole, with the two ends embedded in it.... although
attention looking back may easily decompose the experience, and
distinguish its beginning from its end. p.574-575
[Trabasso and Stein] : the plan unites the past (a desired state) with the
present (an attempt) and the future (the attainment of that state).
3y child: tenseless narrative [storybook: Frog, where are you?]
4y child: temporal sequencing, some actions as relevant to goal
5y child: awareness of action as instrumental toward goal
9y: action-goal relationships are marked increasinlgy
adult: narrative completely captures failure/success of attempts
--> 3y child: "glued to present"
T Trabasso and N.L. Stein. Using goal-plan knowledge to merge the
past with the present and future in narrating events on line. In
M.H. Haith, J.B. Benson, R.J. Roberts, and B.F. Pennington, editors,
The development of future-oriented processes, pages
323–352. University of Chicago Press, 1994.
- TIME AS SUCCESSION
Stimulus Onset Asynchrony expts:
<44s : perceived as simultaneous;
slightliy longer - perception of flicker - no ordering - if
spatially shifted may view as motion.
If one circle is red, next is green, subj perceives a
motion, and the colour changes midway.
Block: while encoding an event, one simultaneously recalls related
preceding events, and anticipates related future events. The
relation ‘e precedes now’ may then be defined operationally as:
‘if I recall event e, it must have taken place before now’, and
analogously for the relation ‘now precedes d’: if d is
anticipated, it must lie in the future.
R.A. Block. Cognitive models of psychological time. Lawrence Erlbaum,
1990.
Michon:
Time is the conscious experiental product of the processes that allow the
(human) organism to adaptively organize itself so that its behaviour
remains tuned to the sequential (i.e. order) relations in its
environment. (Michon [79, p. 40])
[what of subconscious notions of time? as in the SOA expts? ]
Argues that catching a ball is possible based on rate of change of retinal
image, indep of a consc awareness of initial/final position, velocity,
etc.:
many motor skills do not involve explicit time (e.g. in the form of a
clock), which is there to become aware of. Hence, if our conscious
experience of time has a function, it is most likely not that of
facilitating synchronization. So why then do we need the experience
of time? p.11
And suggests that the conscious awareness of time is related to the human
ability for planning
our sense of time derives from being goal–oriented agents p.13
This is the conceptual basis for the formalizations that follow.
What is an event?
Zacks and Tversky:
‘a segment of time at a given location that is conceived by an observer
to have a beginning and an end’.
but not universal - ends may be imprecise (e.g. World War 2), or may have
been interrupted for long periods ("writing a book").
Zacks and Tversky introduce the very interesting hypothesis that mental
representations of events are governed by the ‘equation’
object :: space = event :: time
boundary of event := discontinuity in type of behaviour (e.g. verbal -> social),
body part used, direction of motion, goals, tempo, etc.
Role of granularity - longer events are seen as more intentional (e.g. peace
treaty vs handshake). p.16
The smallest psychologically reified events, on the order of a few
seconds, may be defined primarily in terms of simple physical
changes. For example, think of a person grasping another’s hand, the
hands going up, going down, releasing. Longer events, lasting from
about 10s to 30s, may be defined in relation to some straightforward
intentional act: the events described above, on the time scale
indicated, form a handshake. From a few minutes to a few hours, events
seem to be characterized by plots (i.e. the goals and plans of their
participants) or by socially conventional form of activity. Perhaps the
handshake was part of signing a treaty. On time scales that are long
enough, it may be that events are characterized thematically. In this
example, perhaps the treaty signing was part of an event called a ‘peace
process’. In general, it seems that as the time scale increases, events
become less physically characterized and more defined by the goals,
plans, intentions and traits of their participants. [134, p. 7]
J.M. Zacks and B. Tversky. Event structure in perception and
cognition. Psychological Bulletin, 127(1):3–21, 2001.
When a coarse temporal grain is insufficient to achieve this understanding,
people shift to a finer grain of encoding. It will be noticed that this
picture is completely in line with Marr’s view of object recognition and of
intelligence generally (cf. chapters 5 and 7 of Marr:Vision) p.17
Newton believed that time is a physical entity in itself:
‘Absolute, true and mathematical time, in and of itself, in its own
nature flows equably and without relation to anything external ...’.
Leibniz believed that time is relative in the sense that it is dependent on
the events that occur: no events, no time, and moreover, the structure of
time depends on the structure of events (see [131] and [132] for
discussion).
Russell, in ‘Our knowledge of the external world’ ([95], cf. also [96]), was
concerned with formalizing the latter point of view, as part of a program to
reduce all knowledge of the world to sense data. His construction was later
taken up and somewhat modified by Kamp in [58], and it is this version that
we shall discuss.
[58] H. Kamp. Events, instants and temporal reference. In R. Baeuerle,
U. Egli, and A. von Stechow, editors, Semantics from different points of
view, pages 27–54. Springer Verlag, Berlin, 1979.
[95] B. Russell. Our knowledge of the external world. Allen and Unwin, London,
1914.
[96] B. Russell. On order in time. Proc. Camb. Phil. Soc., 32:216–228, 1936.
[131] G.J. Whitrow. The natural philosophy of time. Clarendon Press, Oxford,
second edition, 1980.
[132] G.J. Whitrow. What is time? Clarendon Press, Oxford, 2003. Abridged
edition of [131].
seven axioms then characterize event structures (all variables universally
quantified) [ only 2 binary predicates: P = precedes; O = overlaps]
(1) P(x, y) => ¬P(y, x)
(2) P(x, y) ^ P(y, z) => P(x, z)
(3) O(x, x)
(4) O(x, y) => O(y, x)
(5) P(x, y) => ¬O(x, y)
(6) P(x, y) ^ O(y, z) ^ P(z, v) => P(x, v)
(7) P(x, y) | O(x, y) | P(y, x)
The last axiom blatantly forces linearity of time, which is somewhat
disappointing, since it seems hard to motivate it independently of
linearity. We could simplify the axioms by defining O(x, y) as ¬P(x, y) ^
¬P(y, x), but this definition has linearity built in. The possibility to
define O(x, y) emphasizes, however, that in this setup only the ‘precedes’
relation is truly primitive. 18
Defines an "instant" as the maximal set of overlapping events.
Formalization [ E = universe of events (undefined)]
* i = subset e IN E s.t. c,d IN i => O(c,d)
* If e IN E but not IN i, then Exists d IN i s.t. ~O(d,e)
?? intuition is that if e is to the Left of the intersection(i), then
must exist a d to the right (else how is the left boundary of i
defined?)
* instants i,j are ordered i3. Walker's Construction of Instant
[notation change, B below is P in vLH, P stands for Past, but overloads P ]
instant = a triple B,C,F [~= back, interior, front of event], s.t.
* E = B U C U F
* B, F nonempty [C empty => instant]
* a IN B, b IN F implies P(a,b)
* if c IN C exists a IN B, b IN F s.t. O(a,c) ^ O(b,c)
BCF is an instant - is the "empty gap between two events" - hence the last
statement - no gap between past and future.
4. Richer languages for events
Is the language containing only the predicates P and O sufficiently rich to
express all possible temporal relationships between events?
predicate B(c, d) (‘c begins before d’) and
E(c, d) (‘c ends before d’)
defines these in terms of Precedes
immediate continguity [ABUTS,MEET]
(1) John pushed the button. Immediately, light flooded the room.
A(a, b) = a abuts b from the left.
Def:
A(a, b) iff P(a, b) ^ ¬Exst c(P(c, b) ^ E(a, c)) ^ ¬Exst d(P(a, d) ^ B(d, b)).
i.e. a and b are E-maximal subsets
rest of chapter analyses two example passages (in French), using Russell's
and Walker's formalisms.
Ch 4: Events Formalized
event predicates need an explicit parameter for time - representeed here by
Reals: structure ( R, <; +,x, 0,1 )
<-- based on axiomatization by Hodges
Event Calculus EC
Ch 5: Logic programming with Time and Events
Speaking intuitively, (R, 0, 1,+, ·,<) is the structure
underlying analytic geometry, or, equivalently, Euclidean geometry. The
language allows one to define polynomials (e.g. xy3 + yz + 5), and the
axioms fix the properties of the operations +, · and determine which polynomials
have real solutions. For us, the most important property of this
structure is that the sets of reals definable in this structure are always of a
very simple kind. For instance, let h(x) be a formula in the language L
containing one free variable x. Now whatever the number of quantifiers in
h(x), the set {x 2 R | (R, 0, 1,+, ·,<) |= h(x)} can always be written
as the union of a finite set of intervals and a finite set of points.
This is a consequence of Tarski’s celebrated theorem on ‘quantifier
elimination for real-closed fields’; the interested reader is referred to
chapter 8 of Hodges [50] for the statement and its proof. Its importance for
us can be seen from the following consideration. Typically, computing with
time and events involves determining the set of t such that HoldsAt(f, t) is
true, where f is a fluent which represents, say, an activity. Now HoldsAt(f,
t) will be characterized by a logic program, which for present purposes can
be equated with a complicated formula in the variable t. Using Tarski’s
theorem one may then show that the set of t such that HoldsAt(f, t) can also
be written as a finite union of intervals and points. This is intuitively
satisfying, since it shows that generally events must have this temporal
profile6.
CAUSALITY
complex examples
7: AKTIONSART
ASPECTUAL CLASSES:
[Ryle 1949]
- accomplishments = change of state which have some "task" associated
- achievement = change of state without a "task"
[Kenny 1963] - ignores Ryle and focuses on diff between state, activity, and
performance.
[Vendler 1957,1967]: four way classification - basis of [Dowty 1979]'s
seminal semantic analysis
[Smith 1991] - adds a 5th class, semelfactives
Carlota Smith, The parameter of aspect, Kluwer 1991
STATES ACTIVITIES ACHIEVEMENTS ACCOMPLISHMENTS
know run recognize paint a picture
believe walk spot/notice make a chair
have swim find/lose deliver a sermon
desire push a cart reach draw a circle
love drive a car die recover from an illness
understand build a house
be happy
from https://www.cs.tcd.ie/Tim.Fernando/E5/hvl.pdf:
Applies the event calculus ideas from to natural language semantics.
Murray Shanahan. Solving the Frame Problem: A Mathematical Investigation
of the Common Sense Law of Inertia. MIT Press, Cambridge (MA), 1997.
A crucial principle in [LH04] is that of inertia. We shall construe this as
a requirement that inertial fluents ' persist forward and backward in time
unless some force is applied on them.
---blurb:
THE PROPER TREATMENT OF EVENTS offers a novel approach to the
semantics of tense and aspect motivated by cognitive considerations.
The book begins by presenting data about the human conceptualization
of time, proposing that planning is important in this regard, and
hence equally for the linguistic encoding of time as tense and aspect.
It then introduces a formal theory of planning, a combination of an
event calculus as developed in Artificial Intelligence with a truth
theory and logic programming techniques. The combined system is
then applied to detailed analyses of tense, grammatical and lexical
aspect, coercion, and different types of nominalizations.
Written accessibly, it is a valuable resource for students and scholars
in theoretical linguists, as well as in philosophy of language, logic,
cognitive science, and computer science.The book is accompanied by
a website at http://staff.science.uva.nl/~michiell providing slides for
instructors and background material for students.
amitabha mukerjee (mukerjee [at-symbol] gmail)
2012 Dec 28