Okay, I'll take a crack at it, although this is basically just a translation of Tyler's comments from the [chat room][1]. Recall that given a topological group (or group like $A_\infty$-space) $A$, we can form the bar complex associated to $A$, $BA$ such that $\Omega BA\simeq A$. $BA$ is what is commonly known as the _delooping_ of $A$. It is a defining characteristic of monoids that we can form this delooping. Moreover, the simplicial object of which this delooping is the geometric realization satisfies a special property often called a _Segal condition_. Intuitively, this condition says that at level $n$, this simplicial object looks like the $n$-fold tensor product (in whatever way tensor product makes sense to us) of the first level of the simplicial object. Note also that we're taking the tensor product relative to the zeroth level of the simplicial object. As a result, if we were interested considering group-objects in some category it would suffice to look at simplicial objects satisfying the Segal condition whose zeroth level is a terminal object (if we don't have this last condition, we would be looking at group-_oid_ objects). 

Now, dualizing, we should say that an object is a comonoid if it admits a _looping_ which we model with the cobar complex (which we can always construct given a comonoidal structure). And the cobar complex satisfies a suitable _coSegal condition_. At first glance this seems kind of silly, since we can ALWAYS loop a space. But I claim that this is precisely because spaces are ALWAYS comonoids! This of course is just by using the diagonal map. Notice however that if we're not working in a concrete category, this may not always be the case. 

So, a coassociative comonoid in $A_\infty$-ring spectra should be precisely the data of a cosimplicial object in $A_\infty$-rings satisfying a coSegal condition. I think that just composing the suspension spectrum functor with the cosimplicial space defining the comonoidal structure on a given space $X$, we can obtain that, using this definition, every space yields a comonoid in spectra. What's more, if this space is an $n$-fold loop space, then this functor can be lifted to $E_n$-ring spectra. 

Now, we haven't touched the issue of cocommutativity at all. For the extent of this answer, we'll just have to take on faith that the right place for cocommutative coalgebras of symmetric monoidal $\infty$-category $C$ to live is $Alg_{E_\infty}(C^{op})$. The motivation for one to believe this is essentially section 5.2 of Lurie's "Higher Algebra" where it is shown that this is the place that the iterated cobar construction lands (producing a Koszul dual monoid to a commutative algebra) and that the iterated bar construction gets us back here. 

So, the point is that, all things considered, I believe that the right definition of a "cocommutative Hopf-algebra in associative ring spectra" is precisely an $E_\infty$-algebra in $(Alg(Spectra)^{op})$. 

Now, how do we know that the suspension spectrum of a loop space satisfies this? Well, first we need to notice that the suspension spectrum functor lands in $Alg(Spectra)$, as expected. Next, we need to notice that it is a monoidal functor with respect to the smash product in both categories, so there is a suitable monoidal op-ification of it (living inside the category of infinity categories) which takes commutative monoids in the opposite category of loop spaces (which, by the way, is ALL OF THEM), to commutative monoids in the opposite category of associative ring spectra.


  [1]: http://chat.stackexchange.com/rooms/9417/homotopy-theory