Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news3.near.net!noc.near.net!paperboy.wellfleet.com!news-feed-1.peachnet.edu!news.duke.edu!agate!howland.reston.ans.net!pipex!uknet!bcc.ac.uk!uu-igor-mac.open.ac.uk!user
From: A.J.Hirst@uk.ac.open (Tony Hirst)
Subject: Re: Lamarckian Evolution
Sender: news@ucl.ac.uk (Usenet News System)
Message-ID: <A.J.Hirst-270195165423@uu-igor-mac.open.ac.uk>
Date: Fri, 27 Jan 1995 16:54:23 GMT
References: <A.J.Hirst-060195122247@uu-igor-mac.open.ac.uk> <3eufpiINN3ojd@rs1.rrz.Uni-Koeln.DE> <A.J.Hirst-160195143550@uu-igor-mac.open.ac.uk> <78774@sdcc12.ucsd.edu>
Organization: HCRL, The Open University, UK
Followup-To: comp.ai.alife
Lines: 103

Before rejoining the fray, i'll make a position statement: i'm trying to
consider the inheritance of acq'd chars (Lam. inh.) in a theoretical sense,
as might be applied to artificially evolved programs, robots etc, rather
than historical biological systems.

> In article <A.J.Hirst-160195143550@uu-igor-mac.open.ac.uk> A.J.Hirst@uk.ac.open (Tony Hirst) writes:

> >> : if we allow for lamarckian inheritance, then we may find that we are
> >> : evolving increasingly brittle children that can survive only in a
> >> : very restricted range of environments;
>  (snip)
> >if by lamarckian inheritance we mean the general term 'the
> >inheritance of acquired characteristics' rather than just use
> >inheritance then:
> >
> >if an individidual is capable of genetically transmitting an encoding of
> >some characteristic it acquired during its lifetime, and it lives in a
> >stable environment, than it will pass on presumably useful traits suited to
> >that environment; if howver, the envt changes over a time period
> >corresponding to a single generation, then the parent will pass on a trait
> >that is adaptive in its envt, but which may be maladaptive for the child's
> >envt.; if the envt is changing even more rapidly, and the parent individual
> >changes to track the envt, then the traits it passes are dependent on the
> >mechanism used to immartalise an acqd char. in the germ line.

> In article <78774@sdcc12.ucsd.edu>, bgrundy@cs.ucsd.edu (Bill Grundy) wrote:

> If I read this correctly, then, you are arguing that Lamarckian
> inheritance would only be a good idea in a stable environment.  That
> argument, however, seems counterintuitive to me.

and me...must have gone wrong somewhere.....

in a stable envt, we would expect a phylogenetic line to converge to some
sort of adapted state anyway, just by using darwinian
operators...learning+baldwin would accelerate this process, lamarckian inh.
even more so...

> Lamarckian evolution should allow the genotype to adjust to changes in
> the environment more quickly than its non-Lamarckian alternative.

agreed - but this is where we have to be clear about what is being acquired
etc.
The first (trivial) is to follow lamarck in arguing for inheritance of
'willed' acquired change (rather than the 'inheritance of mutilation' and
non-adaptive traits - this latter will be harmful to following
generations). I was also making an assumption (as I realise now in trying
to move the goalposts!) relating to an implicit and confused distinction
between the additive acquisition of traits etc, (although i would expect
the traits that get superseded would (genetically) atrophy over time) which
requires extending the genome, and 'replacement acquisition' (ie rewriting
an existing part of the genetic script). If the genome is of fixed length,
then 'genetic forgetting' is required and possibly useful traits may be
lost; if it is extendable, then i think the sys will become more brittle,
'top heavy' and even too specialised to cope with furthewr changes
(although 'forgetting' and atrophy toi reduce genome length may well occur
in this case)- this is a feeling i have, rather than argued belief, but i
think a well designed set of expts would clarify the differences between
the these modes of genomic change.

> Maybe a population which somehow directly passed on acquired traits
> would not be able to cope gracefully with an environment which changes
> too quickly,

if it changes too quickly, i don't think IAC copes at all...

> but surely such a population would do better than some
> other population which lacks the Lamarckian inheritance mechanism.

not if you pass on already outdated traits
 
> After all, when the environment makes a quick, drastic change, real
> (i.e., non-Lamarckian) creatures must adapt in one of three ways: (1)
> by finding an already-existing solution in the current population, (2)
> through mutation and crossover, or (3) via learning and eventual
> incorporation of a new trait into the genome via the Baldwin effect.
> Of these options, the first cannot be relied upon in every situation,
> the second could take an arbitrarily long time, and the third is
> clearly slower than direct inheritance of ontogenically acquired
> traits.
> 
> There might be other ways for a population to aquire a new trait, but
> my point is that what we've been calling Lamarckian inheritance is the
> fastest possible means by which a population can react to a changing
> environment.  Thus, the creation of "brittle individuals" doesn't seem
> to be an argument against the desirability of Lamarckian inheritance.
> 
> Bill Grundy
> -----
> bgrundy@cs.ucsd.edu

When talking of IAC I think we have to consider at least:
1) the mode of inheritance - what acquisitiions are passed on and how does
this affect the genome
2) the relative dynamics of environmental change and trait acquisition...



--------------------------------------------------------------------------
 All opinions etc etc...
--------------------------------------------------------------------------
      | Tony Hirst ("Monty")          | e-mail:  A.J.Hirst@open.ac.uk
       -------------------------------------------------------------------
