An analytial deomp osition of trust in terms of mental - - PowerPoint PPT Presentation

an analyti al de omp osition of trust in terms of mental
SMART_READER_LITE
LIVE PREVIEW

An analytial deomp osition of trust in terms of mental - - PowerPoint PPT Presentation

An analytial deomp osition of trust in terms of mental attitudes (w o rk in p rogress) 1 Rob ert Demolomb e 1 Institut de Reherhe en Info rmatique de T oulouse June 2012 Demolomb e An analytial deomp


slide-1
SLIDE 1 An analyti al de omp
  • sition
  • f
trust in terms
  • f
mental attitudes (w
  • rk
in p rogress) Rob ert Demolomb e 1 1 Institut de Re her he en Info rmatique de T
  • ulouse
June 2012 Demolomb e An analyti al de omp
  • sition
  • f
trust in terms
  • f
mental attitudes (w
  • rk
in p rogress)
slide-2
SLIDE 2 Assumption ( la Castelfran hi et al.): the truster b elieves that if he has some pa rti ula r goal, then this goal will b e rea hed Analysis
  • f
p
  • ssible
supp
  • rts
fo r trust:

Empiri al

truster's
  • bservations

info rmation from trusted sour es

Analyti al

trust in something an b e supp
  • rted
b y trust in
  • ther
things 2
slide-3
SLIDE 3 Obje tive
  • f
this w
  • rk:
systemati analysis
  • f
the p
  • ssible
so ial relationships b et w een the truster and the trustees Metho d:

fo rmalization in mo dal logi

no analysis
  • f
the logi al p rop erties
  • f
the involved mo dalities: b elief, a tion, intention,
  • bligation,
...

trust has the from
  • f
  • nditional
p rop erties 3
slide-4
SLIDE 4 Mo dalities and
  • p
erato rs

φ ⇒ ψ

: φ entails ψ Bel iφ : i b elieves φ Goal iφ : i 's goal is φ A ttempt iφ : i attempts to b ring it ab
  • ut
that φ Int iφ : i 's intention is to b ring it ab
  • ut
that φ Obgφ : it is
  • bligato
ry that φ F
  • rbφ
: it is fo rbidden that φ F
  • rbφ
def

=

Obg¬φ Ask i, jφ : i asks j to b ring it ab
  • ut
that φ Commit i(φ | ψ) : i has
  • mmitted
himself to b ring it ab
  • ut
that φ if ψ holds

φ

: φ holds no w and alw a ys in the future

♦φ

: φ will hold at some moment in the future

♦φ

def

= ¬¬φ

4
slide-5
SLIDE 5 Logi al p rop erties Bel i
  • b
eys a K system Prop erties
  • f
the
  • nditionals
(su ient) (EQUIV) Si ⊢ φ ↔ φ′ et ⊢ ψ ↔ ψ′ , alo rs ⊢ (φ ⇒ ψ) → (φ′ ⇒ ψ′) (PROPG) (φ 1 ∧ φ 2 ⇒ φ 3) → (φ 1 ∧ φ 2 ⇒ φ 1 ∧ φ 3) (TRANS) (φ 1 ⇒ φ 2) ∧ (φ 2 ⇒ φ 3) → (φ 1 ⇒ φ 3) (DIST) (φ 1 ∧ (φ 1 ⇒ φ 2) ∧ ψ) ⇒ (φ 2 ∧ ψ) (MA TIMP) (φ ⇒ ψ) → (φ → ψ) 5
slide-6
SLIDE 6 Initial fo rm
  • f
trust i 's goal: to rea h a state
  • f
aairs Example: i b elieves that if he has no ash (¬φ) and his goal is to get ash (♦φ), then he will get ash (♦φ) (F1) Bel i(¬φ ∧ Goal i♦φ ⇒ ♦φ) 6
slide-7
SLIDE 7 Initial fo rm
  • f
trust i 's goal: to rea h a state
  • f
aairs Example: i b elieves that if he has no ash (¬φ) and his goal is to get ash (♦φ), then he will get ash (♦φ) (F1) Bel i(¬φ ∧ Goal i♦φ ⇒ ♦φ) i 's goal: to maintain a state
  • f
aairs Example: i b elieves that if his a r w
  • ks
w ell (φ) and his goal is that it still w
  • rks
w ell (φ), then it still w
  • rks
w ell (φ) (M1) Bel i(φ ∧ Goal iφ ⇒ φ) 7
slide-8
SLIDE 8 Analysis
  • f
trust supp
  • rt
to rea h If there is some agent j who holds some p rop ert y Prop( j, φ) su h that: (F2) Bel i((¬φ ∧ Goal i♦φ ⇒ ∃ jProp( j, φ)) ∧ (∃ jProp( j, φ) ⇒ ♦φ)) (F2) is a supp
  • rt
fo r (F1) (F1) Bel i(¬φ ∧ Goal i♦φ ⇒ ♦φ) b e ause (F2) entails (F1) 8
slide-9
SLIDE 9 Analysis
  • f
trust supp
  • rt
to rea h If there is some agent j who holds some p rop ert y Prop( j, φ) su h that: (F2) Bel i((¬φ ∧ Goal i♦φ ⇒ ∃ jProp( j, φ)) ∧ (∃ jProp( j, φ) ⇒ ♦φ)) (F2) is a supp
  • rt
fo r (F1) (F1) Bel i(¬φ ∧ Goal i♦φ ⇒ ♦φ) b e ause (F2) entails (F1) to maintain If there is no agent who holds some p rop ert y Prop′( j, φ) su h that: (M2) Bel i((φ ∧ Goal iφ ⇒ ¬∃ jProp′( j, φ)) ∧ (¬∃ jProp′( j, φ) ⇒ φ)) (M2) is a supp
  • rt
fo r (M1) i 's assumption: the
  • nly
w a y to hange φ is that there is an agent j su h that Prop′( j, φ) 9
slide-10
SLIDE 10 T rust in abilit y (F2) Bel i((¬φ ∧ Goal i♦φ ⇒ ∃ jProp( j, φ)) ∧ (∃ jProp( j, φ) ⇒ ♦φ)) Abilit y Able jφ def

=

A ttempt jφ ⇒ ♦φ T rust in abilit y: Bel i Able jφ Prop 1( j, φ) def

=

A ttempt jφ ∧ ( A ttempt jφ ⇒ ♦φ) 10
slide-11
SLIDE 11 T rust in abilit y (F2) Bel i((¬φ ∧ Goal i♦φ ⇒ ∃ jProp( j, φ)) ∧ (∃ jProp( j, φ) ⇒ ♦φ)) Abilit y Able jφ def

=

A ttempt jφ ⇒ ♦φ T rust in abilit y: Bel i Able jφ Prop 1( j, φ) def

=

A ttempt jφ ∧ ( A ttempt jφ ⇒ ♦φ) i 's b elief:

∃ j(

A ttempt jφ ∧ Able jφ) ⇒ ♦φ If Prop( j, φ) is Prop 1( j, φ) , (F2) holds 11
slide-12
SLIDE 12 T rust in abilit y (M2) Bel i((φ ∧ Goal iφ ⇒ ¬∃ jProp′( j, φ)) ∧ (¬∃ jProp′( j, φ) ⇒ φ)) Abilit y Prop′ 1( j, φ) def

=

A ttempt j¬φ ∧ ( A ttempt j¬φ ⇒ ♦¬φ) 12
slide-13
SLIDE 13 T rust in abilit y (M2) Bel i((φ ∧ Goal iφ ⇒ ¬∃ jProp′( j, φ)) ∧ (¬∃ jProp′( j, φ) ⇒ φ)) Abilit y Prop′ 1( j, φ) def

=

A ttempt j¬φ ∧ ( A ttempt j¬φ ⇒ ♦¬φ) Logi al p rop erties:

⊢ ∃ jProp′

1( j, φ) → ♦¬φ

⊢ φ → ¬∃ jProp′

1( j, φ) i 's b elief:

¬∃ j(

A ttempt j¬φ ∧ Able j¬φ) ⇒ φ If Prop′( j, φ) is Prop′ 1( j, φ) , (M2) holds 13
slide-14
SLIDE 14 A tive j 's intention triggers j 's a tion A tive jφ def

=

Int jφ ⇒ A ttempt jφ Prop 2( j, φ) def

=

Int jφ ∧ ( Int jφ ⇒ A ttempt jφ) ∧ Able jφ 14
slide-15
SLIDE 15 A tive j 's intention triggers j 's a tion A tive jφ def

=

Int jφ ⇒ A ttempt jφ Prop 2( j, φ) def

=

Int jφ ∧ ( Int jφ ⇒ A ttempt jφ) ∧ Able jφ Logi al p rop ert y:

⊢ ∃ jProp

2( j, φ) → ♦φ i 's b elief:

∃ j(

Int jφ ∧ A tive jφ ∧ Able jφ) ⇒ ♦φ 15
slide-16
SLIDE 16 A tive Prop′ 2( j, φ) def

=

Int j¬φ ∧ ( Int j¬φ ⇒ A ttempt j¬φ) ∧ Able j¬φ Logi al p rop ert y:

⊢ ∃ jProp′

2( j, φ) → ♦¬φ i 's b elief:

¬∃ j(

Int j¬φ ∧ A tive j¬φ ∧ Able j¬φ) ⇒ φ 16
slide-17
SLIDE 17 Intention adoption No rms fulllment T
  • a hieve
If j b elieves that he is
  • bliged
to do something, then he intends to do that thing Ob ey jφ def

=

Bel j ObgInt jφ ⇒ Int jφ Prop
  • 3. 1(
j, φ) def

=

Bel j ObgInt jφ ∧ ( Bel j ObgInt jφ ⇒ Int jφ)∧ A tive jφ ∧ Able jφ 17
slide-18
SLIDE 18 Intention adoption No rms fulllment T
  • a hieve
If j b elieves that he is
  • bliged
to do something, then he intends to do that thing Ob ey jφ def

=

Bel j ObgInt jφ ⇒ Int jφ Prop
  • 3. 1(
j, φ) def

=

Bel j ObgInt jφ ∧ ( Bel j ObgInt jφ ⇒ Int jφ)∧ A tive jφ ∧ Able jφ Logi al p rop ert y:

⊢ ∃ jProp

  • 3. 1(
j, φ) → ♦φ i 's b elief:

∃ j(

ObgInt jφ ∧ Ob ey jφ ∧ A tive jφ ∧ Able jφ) ⇒ ♦φ 18
slide-19
SLIDE 19 No rms fulllment and institutional p
  • w
er If i asks j to b ring it ab
  • ut
that φ, then j b elieves that it is
  • bligato
ry that he adopts the intention to b ring it ab
  • ut
that φ Example: p
  • li eman
i asks j to stop his a r InstP
  • w
er i, jφ def

=

Ask i, jφ ⇒ Bel j ObgInt jφ Prop
  • 4. 1(
j, φ) def

=

Ask i, jφ ∧ ( Ask i, jφ ⇒ Bel j ObgInt jφ)∧ Ob ey jφ ∧ A tive jφ ∧ Able jφ 19
slide-20
SLIDE 20 No rms fulllment and institutional p
  • w
er If i asks j to b ring it ab
  • ut
that φ, then j b elieves that it is
  • bligato
ry that he adopts the intention to b ring it ab
  • ut
that φ Example: p
  • li eman
i asks j to stop his a r InstP
  • w
er i, jφ def

=

Ask i, jφ ⇒ Bel j ObgInt jφ Prop
  • 4. 1(
j, φ) def

=

Ask i, jφ ∧ ( Ask i, jφ ⇒ Bel j ObgInt jφ)∧ Ob ey jφ ∧ A tive jφ ∧ Able jφ Logi al p rop ert y:

⊢ ∃ jProp

  • 4. 1(
j, φ) → ♦φ i 's b elief:

∃ j(

Ask i, jφ ∧ Ob ey jφ ∧ InstP
  • w
er i, jφ ∧ A tive jφ ∧ Able jφ) ⇒ ♦φ 20
slide-21
SLIDE 21 No rms fulllment T
  • maintain
Prop′
  • 3. 1(
j, φ) def

=

Bel j ObgInt j¬φ ∧ ( Bel j ObgInt j¬φ ⇒ Int j¬φ)∧ A tive j¬φ ∧ Able j¬φ 21
slide-22
SLIDE 22 No rms fulllment T
  • maintain
Prop′
  • 3. 1(
j, φ) def

=

Bel j ObgInt j¬φ ∧ ( Bel j ObgInt j¬φ ⇒ Int j¬φ)∧ A tive j¬φ ∧ Able j¬φ Logi al p rop ert y:

⊢ ∃ jProp′

  • 3. 1(
j, φ) → ♦¬φ i 's b elief:

¬∃ j(

Bel j ObgInt j¬φ ∧ Ob ey j¬φ ∧ A tive j¬φ ∧ Able j¬φ) ⇒ φ

¬∃ jProp′

  • 3. 1(
j, φ) : no agent who fullls the no rms b elieves that ObgInt j¬φ Prop′′
  • 3. 1(
j, φ) def

=

Bel j F
  • rbInt
j¬φ ∧ ( Bel j F
  • rbInt
j¬φ ⇒ Int j¬φ)∧ A tive j¬φ ∧ Able j¬φ

¬∃ jProp′

  • 3. 1(
j, φ) : no agent who violates the no rms b elieves that F
  • rbInt
j¬φ 22
slide-23
SLIDE 23 Contra t T
  • a hieve
If i asks to a taxi driver j to
  • mmit
himself to b ring i at the airp
  • rt
in a
  • ntext
where i
  • mmits
himself to pa y the taxi driver, then the taxi driver j adopts the intention to b ring i at the airp
  • rt
Contra t i, j(φ, ψ) def

=

Ask i, j Commit j(φ | Commit iψ) ⇒ Int jφ Prop
  • 3. 2(
j, φ) def

=

Ask i, j Commit j(φ | Commit iψ) ∧

(

Ask i, j Commit j(φ | Commit iψ) ⇒ Int jφ)∧ A tive jφ ∧ Able jφ 23
slide-24
SLIDE 24 Contra t T
  • a hieve
If i asks to a taxi driver j to
  • mmit
himself to b ring i at the airp
  • rt
in a
  • ntext
where i
  • mmits
himself to pa y the taxi driver, then the taxi driver j adopts the intention to b ring i at the airp
  • rt
Contra t i, j(φ, ψ) def

=

Ask i, j Commit j(φ | Commit iψ) ⇒ Int jφ Prop
  • 3. 2(
j, φ) def

=

Ask i, j Commit j(φ | Commit iψ) ∧

(

Ask i, j Commit j(φ | Commit iψ) ⇒ Int jφ)∧ A tive jφ ∧ Able jφ Logi al p rop ert y:

⊢ ∃ jProp

  • 3. 2(
j, φ) → ♦φ i 's b elief:

∃ j(

Ask i, j Commit j(φ | Commit iψ) ∧ Contra t i, j(φ, ψ) ∧ A tive jφ ∧ Able jφ) ⇒ ♦φ 24
slide-25
SLIDE 25 T
  • maintain
Prop′
  • 3. 2(
j, φ) def

= ∃ k(

Ask j, k Commit k(¬φ | Commit jψ) ∧

(

Ask j, k Commit k(¬φ | Commit jψ) ⇒ Int k¬φ)∧ A tive k¬φ ∧ Able k¬φ Example: i 's goal is not b e killed and j asks to some maa memb er to kill i 25
slide-26
SLIDE 26 T
  • maintain
Prop′
  • 3. 2(
j, φ) def

= ∃ k(

Ask j, k Commit k(¬φ | Commit jψ) ∧

(

Ask j, k Commit k(¬φ | Commit jψ) ⇒ Int k¬φ)∧ A tive k¬φ ∧ Able k¬φ Example: i 's goal is not b e killed and j asks to some maa memb er to kill i Logi al p rop ert y:

⊢ ∃ jProp′

  • 3. 2(
j, φ) → ♦¬φ i 's b elief:

¬∃ jProp′

  • 3. 2(
j, φ) ⇒ φ 26
slide-27
SLIDE 27 Altruism If j b elieves that i 's goal is φ, then j adopts the intention to b ring it ab
  • ut
that φ Example: i is an
  • ld
man who w ants to nd some help to ross the road and j is a w a re
  • f
i 's goal Altruis j, iφ def

=

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int jφ Prop
  • 3. 3(
j, φ) def

=

Bel j(¬φ ∧ Goal i♦φ)∧

(

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int jφ)∧ A tive jφ ∧ Able jφ 27
slide-28
SLIDE 28 Altruism If j b elieves that i 's goal is φ, then j adopts the intention to b ring it ab
  • ut
that φ Example: i is an
  • ld
man who w ants to nd some help to ross the road and j is a w a re
  • f
i 's goal Altruis j, iφ def

=

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int jφ Prop
  • 3. 3(
j, φ) def

=

Bel j(¬φ ∧ Goal i♦φ)∧

(

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int jφ)∧ A tive jφ ∧ Able jφ Logi al p rop ert y:

⊢ ∃ jProp

  • 3. 3(
j, φ) → ♦φ i 's b elief:

∃ j(

Bel j(¬φ ∧ Goal i♦φ) ∧ Altruis j, iφ ∧ A tive jφ ∧ Able jφ) ⇒ ♦φ i b elieves that if there is some altruist agent, his goal φ will b e rea hed 28
slide-29
SLIDE 29 P erversion P ervert j, iφ def

=

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int j¬φ j intends to p revent i to rea h his goal Prop′
  • 3. 3(
j, φ) def

=

Bel j(¬φ ∧ Goal i♦φ)∧

(

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int j¬φ)∧ A tive j¬φ ∧ Able j¬φ 29
slide-30
SLIDE 30 P erversion P ervert j, iφ def

=

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int j¬φ j intends to p revent i to rea h his goal Prop′
  • 3. 3(
j, φ) def

=

Bel j(¬φ ∧ Goal i♦φ)∧

(

Bel j(¬φ ∧ Goal i♦φ) ⇒ Int j¬φ)∧ A tive j¬φ ∧ Able j¬φ Logi al p rop ert y:

⊢ ∃ jProp′

  • 3. 3(
j, φ) → ♦¬φ i 's b elief:

¬∃ j(

Bel j(¬φ∧ Goal i♦φ)∧ P ervert j, iφ∧ A tive j¬φ∧ Able j¬φ) ⇒ ♦φ i b elieves that if there is no p ervert agent the state
  • f
aairs φ will b e maintained 30
slide-31
SLIDE 31 Summa ry Initial trust denition:

to rea h a state
  • f
aairs

to maintain a state
  • f
aairs Analysis
  • f
p
  • ssible
trust supp
  • rts:

A tion and abilit y

Intention and "a tiveness"

Obligation and
  • b
edien e (no rms fulllment)

Request and
  • ntra t
fulllment (mutual interest)

Altruism Dualit y : to rea h vs to maintain

there is a "go
  • d"
agent

there is no "bad" agent 31
slide-32
SLIDE 32 F
  • rmalization
The p rop
  • sed
axiomati
  • f
  • nditionals
is su ient to p rove that ea h p rop ert y Prop 1 , Prop 2 , Prop 3 , ... logi ally entails Prop 32